Hash power network service anycast method and system

Abstract

The application provides a computing power network service anycast method and system, and the steps of the method include: a gateway forwards a service request to a controller; the controller calculates the transmission cost and propagation cost of the gateway to each cloud node and edge node based on the transmission distance and bandwidth of the gateway to the cloud node and the edge node, calculates a network transmission cost matrix based on the transmission cost and propagation cost, calculates an arrival rate matrix based on the forwarding probability weight of the gateway to the cloud node and the edge node, calculates a target anycast forwarding weight strategy matrix by using a Sinkhorn algorithm based on the arrival rate matrix and the network transmission cost matrix, and the target anycast forwarding weight strategy matrix includes the weight value of each cloud node and edge node; the number of positions of the cloud node and the edge node in a queuing queue is determined based on the weight value of the cloud node and the edge node, the queuing queue is constructed, and the processing task corresponding to the service request is distributed to the corresponding cloud node or edge node based on the queuing queue.

Description

Technical Field

[0001] This invention relates to the field of computing power network technology, and in particular to a method and system for anycasting computing power network services. Background Technology

[0002] In the information age, globally interconnected mobile computing has deeply permeated every aspect of human life. Simultaneously, computing models are evolving, from internet-centric cloud computing to edge computing that enables localized business loops and agile intelligence, and further to the future synergy between cloud-edge computing capabilities and end-device platforms. The computing model is developing towards a ubiquitous architecture with multi-level cloud-edge-end deployment. This necessitates expanding end-to-end communication to enhance high-performance computing applications (e.g., emergency intelligent applications), machines and mobile objects (e.g., autonomous vehicles and drones), content distribution and streaming services, and emphasizing the ability to virtualize and replicate any service to improve efficiency and redundancy.

[0003] Current network routing and forwarding systems designed for the Internet are primarily end-to-end communication designs between single-destination nodes. Network devices route data packets carrying application layer requests to a single peer network node based on the destination address. While load balancers placed at locations such as public cloud entry points can balance the load across homogeneous service instances, this system limits the potential for coordinated services between cloud-edge, edge-edge, and multi-cloud environments. Anycast services allow application service replicas in different locations to share a common service identifier and forward service requests to one of multiple replicas according to certain strategies. This provides users with enhanced quality of experience and offers service providers several advantages (e.g., load balancing between replicas, robustness against DDoS attacks). However, a key challenge remains how to coordinate the differences among multiple service replica instances and design service forwarding mechanisms and forwarding decision calculation methods to improve overall service performance while achieving complete decoupling of service and address locations. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method and system for anycasting computing power network services to eliminate or improve one or more defects existing in the prior art.

[0005] One aspect of the present invention provides a method for anycasting computing power network services. This method is based on a pre-built computing power network, which includes a gateway, a controller, multiple cloud nodes, and multiple edge nodes. The steps of the method include:

[0006] Based on the gateway receiving the service request, the gateway forwards the service request to the controller;

[0007] The controller acquires the working status of each cloud node and edge node in real time, and constructs the cloud nodes and edge nodes into a node set;

[0008] Based on the transmission distance and bandwidth between the gateway and the cloud nodes and edge nodes, calculate the transmission cost and propagation cost from the gateway to each cloud node and edge node, and calculate the network transmission cost matrix based on the transmission cost and propagation cost;

[0009] The arrival rate matrix is ​​calculated based on the forwarding probability weights of the gateway, cloud nodes, and edge nodes;

[0010] Based on the arrival rate matrix and the network transmission cost matrix, the Sinkhorn algorithm is used to calculate the target anycast forwarding weight strategy matrix, which includes the weight values ​​of each cloud node and edge node.

[0011] The number of cloud nodes and edge nodes in the queuing queue is determined based on the weight values ​​of cloud nodes and edge nodes, and the queuing queue is constructed. Based on the queuing queue, the processing tasks corresponding to the service requests are distributed to the corresponding cloud nodes or edge nodes.

[0012] By adopting the above scheme, this scheme achieves an organic combination of centralized control and distributed operation through the separation of the control plane and data plane of software-defined networking (SDN). In the complex environment of cross-layer nodes, the computing power network service anycast forwarding architecture fully utilizes the global optimization capability of centralized control and the flexibility of the queue forwarding mechanism based on weighted strategy in the distributed data plane. This enables the system to perform fine-grained balanced scheduling of burst traffic and traffic fluctuations from the perspective of global optimization, while ensuring stable service quality under high load conditions and improving resource utilization.

[0013] In some embodiments of the present invention, the method includes the following steps: when the controller receives a service request for the first time, the controller parses the service type corresponding to the service request, calculates the network transmission cost matrix and arrival rate matrix corresponding to the service type, and calculates the queuing queue corresponding to the service type.

[0014] When the controller receives a service request again, it parses the service type corresponding to the service request and determines whether there is a queue for that service type.

[0015] If it exists, the processing task corresponding to the service request will be directly distributed to the corresponding cloud node or edge node based on the existing queue.

[0016] If it does not exist, then construct a queue corresponding to the service type.

[0017] In some embodiments of the present invention, in the step of sending the processing task corresponding to the service request to the corresponding cloud node or edge node based on the queue, the cloud node or edge node at the head of the queue is used as the sending cloud node or edge node, and the cloud node or edge node is moved to the tail of the queue to update the queue.

[0018] In some embodiments of the present invention, in the step of calculating the transmission cost and propagation cost from the gateway to each cloud node and edge node based on the transmission distance and bandwidth between the gateway and the cloud node and the edge node, and calculating the network transmission cost matrix based on the transmission cost and propagation cost:

[0019] Based on the pre-set data packets, the amount of input and output data transmitted from the gateway to each cloud node and edge node is used to calculate the transmission cost based on the input and output data amounts and bandwidth; the time consumed by the pre-set data packets from the gateway to each cloud node and edge node is used as the propagation cost.

[0020] Based on the transmission cost and propagation cost, the gateway transmits the cost to each cloud node and edge node, and constructs a network transmission cost matrix.

[0021] In some embodiments of the present invention, in the step of calculating the transmission cost based on the input data volume and output data volume transmitted from the gateway to each cloud node and edge node based on a preset data packet, and the bandwidth, the transmission cost is calculated based on the following formula:

[0022]

[0023] in, This represents the transmission cost value from router i in the gateway to the cloud node and edge node j; This represents the pre-set amount of input data for service k, transmitted from router i of the gateway to the cloud node and edge node j. This indicates the pre-configured amount of data packets for service k that are transmitted from router i of the gateway to the cloud node and edge node j. This indicates the bandwidth used to transmit pre-configured data packets from router i in the gateway to the cloud node and edge node j.

[0024] In some embodiments of the present invention, in the step of calculating the target anycast forwarding weight strategy matrix using the Sinkhorn algorithm based on the arrival rate matrix and the network transmission cost matrix, the arrival rate matrix and the network transmission cost matrix are used as two diagonal matrices of the Sinkhorn algorithm, and an objective function is constructed based on the arrival rate matrix and the network transmission cost matrix. The final objective function value is determined by iteratively performing the Sinkhorn algorithm to obtain the target anycast forwarding weight strategy matrix.

[0025] In some embodiments of the present invention, in the step of constructing an objective function based on the arrival rate matrix and the network transmission cost matrix, the objective function is constructed as follows:

[0026]

[0027] in, Represents the objective function value. Let the arrival rate matrix represent the arrival rate for service k. This represents the network transmission cost matrix for service k. This indicates the calculation of the inner product of matrices. This indicates minimization.

[0028] In some embodiments of the present invention, in the step of calculating the arrival rate matrix based on the forwarding probability weights of the gateway, cloud node, and edge node, the arrival rate matrix is ​​calculated based on the following formula:

[0029]

[0030] in, Represents the arrival rate matrix. Represents the rate vector. This represents the forwarding strategy matrix.

[0031] In some embodiments of the present invention, in the step of calculating the network transmission cost matrix based on the transmission cost and propagation cost, the transmission cost value transmitted from the gateway router to each cloud node and edge node is calculated, and the propagation cost value transmitted from the gateway router to each cloud node and edge node is measured by a preset data packet for service k. For each cloud node or edge node, the sum of the transmission cost value and the propagation cost value is calculated as the value at each position in the network transmission cost matrix, and the network transmission cost matrix is ​​constructed.

[0032] A second aspect of the present invention also provides a computing power network service anycast system, the system including a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implementing the steps of the method described above.

[0033] A third aspect of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the aforementioned computing power network service anycast method.

[0034] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the text, or may be learned by practice of the invention. The objects and other advantages of the invention will become apparent from the description and the accompanying drawings.

[0035] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0036] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.

[0037] Figure 1 This is a schematic diagram illustrating one implementation of the computing power network service anycast method of this scheme;

[0038] Figure 2 A schematic diagram of the architecture of the computing power network service anycast method in this scheme;

[0039] Figure 3 A schematic diagram illustrating the process of updating the anycast forwarding queue for the computing power network service in this solution;

[0040] Figure 4 This is a schematic diagram of the Packet_in message processing flow in the control plane of the computing power network service in this solution.

[0041] Figure 5 This is a schematic diagram of the algorithm for calculating the service anycast strategy based on optimal transmission in this scheme. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0043] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0044] Introduction to existing technologies:

[0045] Existing Option 1:

[0046] Existing Solution 1 relates to the field of computing power scheduling. To address the problem that static rule-based routing cannot adapt to dynamic environmental changes, this computing power routing method includes: acquiring first computing power resource information and task characteristic information of the computing task to be processed; obtaining a probability distribution for forwarding the computing task to be processed to various neighboring nodes based on the task characteristic information and the first computing power resource information; sampling from the probability distribution to obtain a target forwarding node; and forwarding the task characteristic information to the target forwarding node. This invention can adapt to dynamic environmental changes in the computing power scheduling system, improve the resource utilization and security of the computing power scheduling system, and optimize the performance of the computing power scheduling system.

[0047] Existing Option 2:

[0048] Existing Solution 2 aims to achieve efficient and reliable allocation of computing resources and task scheduling through the learning and decision-making of intelligent agents. This method includes: performing network resource awareness; selecting computing nodes based on task requirements and node evaluation information; using a controller based on graph neural networks and incorporating deep reinforcement learning agents to determine the routing path from the starting node to the destination node; forwarding the determined routing path to nodes in the computing network and updating the state information of the nodes and the network. This invention is based on reinforcement learning algorithms and defines states, actions, and reward functions through graph neural networks, enabling it to learn the optimal routing strategy from the state space. Experimental verification shows that this invention can effectively allocate more computing tasks and maintain stability and reliability under different network topologies and task load conditions.

[0049] The technical disadvantages of the existing solution 1:

[0050] (1) Lack of dynamic load balancing capability and poor policy stability: This computing power routing method only allocates tasks based on node probability distribution sampling and fails to fully consider the real-time load information of service instances. This approach may cause some nodes to continue receiving requests under high load, reducing the overall resource utilization efficiency and easily causing local congestion.

[0051] (2) Lack of a service quality guarantee mechanism: This method mainly relies on static computing resource information and task characteristic information in forwarding decisions, and lacks detailed service quality (QoS) control measures. It is unable to achieve a balance between service quality and resource allocation, thus making it difficult to meet the service needs of multi-level and complex application scenarios.

[0052] (3) It cannot adapt to the service coordination needs across multiple nodes: This method mainly focuses on the scheduling optimization of single computing resources and lacks the ability to coordinate and process resources at the edge, cloud and multi-level, making it difficult to adapt to the effective collaboration of computing resources across cloud edge, edge and multi-cloud environments.

[0053] The technical disadvantages of the existing solution 2:

[0054] (1) High computational and resource overhead: The introduction of deep reinforcement learning and graph neural networks has significantly increased computational complexity and resource overhead, especially in multi-layered, large-scale network topologies. This high overhead may affect the real-time performance of the system and is not conducive to responding quickly to task requirements in dynamic computing environments.

[0055] (2) Model training relies on a large amount of data: Training reinforcement learning models requires a large amount of task allocation and routing decision data to learn effective policies. However, in actual network deployment, collecting enough data may lead to latency, or insufficient data may affect the generalization and deployment performance of the model.

[0056] (3) Difficult to cope with rapidly changing network conditions: This method relies on the training of deep reinforcement learning models, so the model update speed is usually slow and it is difficult to adapt to rapidly changing network conditions and task requirements in a timely manner, resulting in low adaptability of real-time routing decisions.

[0057] (4) Low transparency in path selection: Due to the "black box" nature of reinforcement learning models and graph neural networks, it is difficult to explain their specific decision-making paths and selection mechanisms. This lack of transparency may affect the management and maintenance of service operators during the deployment process and increase the complexity of troubleshooting routing faults.

[0058] (5) Limited support for selective broadcast across multiple nodes: This method focuses on task routing path selection and resource allocation, and lacks support for multi-level coordination and selective broadcast forwarding of service instances in the computing power network, which may make it difficult to meet the load balancing requirements of different service instances.

[0059] In summary, existing technologies lack the ability to dynamically schedule service instances with fine granularity. As computing resources are increasingly embedded in network infrastructure, users increasingly expect efficient, low-latency services, rather than focusing solely on specific physical locations. Traditional network architectures, primarily based on end-to-end communication, struggle to adapt to the dynamic scheduling needs of different service instances. Because the location, load, and user requests of service instances often change dynamically, current load balancers and static routing mechanisms cannot flexibly allocate resources, leading to resource waste and unstable service quality. Furthermore, existing designs for anycast forwarding of computing power services (sometimes referred to as computing power routing) often use a single-destination service instance approach for processing service requests on the computing power network access router within scheduling slots, making it difficult to perform fine-grained load balancing for the varying processing capabilities of different service instances.

[0060] Existing technologies suffer from insufficient resource coordination across multiple node layers: Computing networks typically contain multiple layers of resource nodes, including cloud and edge nodes, with significant differences in computing and storage capabilities, geographical distribution, and other aspects. However, current network architectures lack collaborative scheduling mechanisms across these layers, making it difficult for edge nodes' computing resources to fully collaborate with cloud nodes. This lack of collaboration not only increases the burden on individual nodes but also limits the efficiency of global resource utilization, ultimately reducing the overall quality of service.

[0061] The mismatch between network and computing load metrics in existing technologies leads to unstable broadcast selection strategies: While introducing dynamic broadcast selection strategies based on real-time load and network conditions can optimize service response, the indirect processing time metrics (such as CPU utilization) used in computing resource load assessment do not perfectly match actual network and computing conditions. Furthermore, due to the delay in updating computing load information, broadcast selection strategies in such systems are often overly greedy and unstable, resulting in poor adaptation of computing resources and the network, thus impairing service stability and user experience.

[0062] like Figure 1 and 2 As shown, this invention proposes a method for anycasting computing power network services. This method is based on a pre-built computing power network, which includes a gateway, a controller, multiple cloud nodes, and multiple edge nodes. The steps of this method include:

[0063] In the actual implementation process, multiple service replica instances of the same service are concurrently deployed in the network for the computing power network, and different service instances can be deployed on different computing nodes at the same time, such as edge nodes (EN) and cloud nodes (CN), and have different request processing capabilities based on the amount of computing resources and network resources occupied by the instances.

[0064] Step S100: Based on the gateway receiving the service request, the gateway forwards the service request to the controller;

[0065] In practical implementation, the computing network (CPN) access router located near the user (such as a mobile edge computing site) mainly has three functions:

[0066] 1. Request Statistics and Reporting: By collecting statistical information on user service requests, the access router gateway can monitor key data such as user access frequency, request type, and request processing time in real time. This statistical information is periodically reported to the centralized controller so that the controller can make global optimization decisions, such as load balancing and service migration. In addition, the access router can implement local caching based on different service request traffic to improve access performance.

[0067] 2. Anycast Forwarding Queue Processing: The access router gateway determines the next-hop target for service requests based on the anycast forwarding queue received from the controller. During the forwarding process, it adopts a circular forwarding queue mechanism, which allows service requests to be flexibly distributed among different computing nodes, thereby improving the concurrency and response speed of service processing.

[0068] 3. Request Encapsulation and Forwarding: In some cases, access routers and gateways need to encapsulate the original service requests, especially when forwarding requests between different network domains. After encapsulating the external header, the access router forwards the request to the egress router, ensuring that the request can be effectively transmitted across different network environments.

[0069] The computing network egress routers deployed near the computing nodes mainly have two functions:

[0070] 1. Decapsulation and Forwarding: The primary responsibility of the egress router is to decapsulate encapsulated requests from access routers, restoring the original request information. The decapsulated request is then forwarded to the original access router based on the service type and topology, ensuring that the service instance can correctly receive and process the request.

[0071] 2. Service Information Reporting: In addition to forwarding data, the egress router needs to collect service information deployed on the compute nodes, such as service type, processing capacity, and service instance load, and report this information to the controller. This data helps the controller perform dynamic resource scheduling and service instance management, achieving globally optimal resource allocation.

[0072] Step S200: The controller acquires the working status of each cloud node and edge node in real time, and constructs the cloud nodes and edge nodes into a node set;

[0073] In its implementation, this solution uses OpenFlow and the Ryu controller to implement a computing power network service anycast forwarding system, and incorporates key features such as CPN topology awareness, service ID storage, and matching. The Locust stress testing tool is used to evaluate service processing capacity and assign an anycast IP to each service ID.

[0074] Step S300: Based on the transmission distance and bandwidth between the gateway and the cloud nodes and edge nodes, calculate the transmission cost and propagation cost from the gateway to each cloud node and edge node, and calculate the network transmission cost matrix based on the transmission cost and propagation cost.

[0075] Step S400: Calculate the arrival rate matrix based on the forwarding probability weights of the gateway, cloud node, and edge node;

[0076] Step S500: Based on the arrival rate matrix and the network transmission cost matrix, the Sinkhorn algorithm is used to calculate the target anycast forwarding weight strategy matrix, which includes the weight values ​​of each cloud node and edge node.

[0077] In practical implementation, due to the limited programmability of the OpenFlow data plane, this solution adopts a control plane application record written in RYU to maintain the service anycast forwarding table for different CPN access routers and execute forwarding queue updates to implement the service anycast mechanism. The service anycast forwarding entry class contains the following entry attributes:

[0078] 1. The software switch data channel ID (dpid) to which the forwarding entry belongs;

[0079] 2. The service ID to which the forwarded entry belongs;

[0080] 3. The transport layer protocol type of the service matched by this service forwarding entry;

[0081] 4. A list of available service instance IDs for the services matched by the forwarding entry of this service;

[0082] 5. The forwarding strategy currently used for the forwarding entries in this service;

[0083] 6. The latest service forwarding destination ID (newest_target) set for this service forwarding entry;

[0084] 7. Indicates whether the current entry needs to update the tag indicator (the default is that it needs to be updated);

[0085] 8. Record the software switch flow table matching entry corresponding to this forwarding;

[0086] 9. Record the matching entry in the reverse flow table of the software switch corresponding to this forwarding;

[0087] 10. Record the moment when the flow table was last triggered based on the Openflow packet_in message for this forwarding entry;

[0088] 11. Other related settings for forwarding entries in this service;

[0089] 12. The forwarding queue currently being used by the forwarding entries for this service;

[0090] The key service forwarding queue update generation function for the service anycast forwarding entry class, such as... Figure 3 As shown in the diagram. In this function, the forwarding policy `forwarding_policy` is first input. This policy element represents the probability weight of forwarding a service request to the corresponding service instance. Then, a new forwarding queue `forwarding_queue` is created. The product of the forwarding policy and the number of instances is rounded to generate the policy. This policy element indicates the number of different destination service instances in the forwarding queue. The loop stops when the sum of the policy elements is less than a small value, such as 10^(-5). Otherwise, the policy elements are iterated through, and each policy element is decremented by 1, and the service instance indicated by the policy element is added to a temporary queue `temp`. This process continues until all policy elements have been iterated through. Finally, the `temp` queue is shuffled and added to the forwarding queue `forwarding_queue`. These operations ensure that even if the forwarding queue mechanism cannot be implemented on the data plane, implementing it on the control plane can reduce the imbalance caused by service requests with the same forwarding decision cycle being concurrently sent to the same destination instance, achieving a more balanced service forwarding effect.

[0091] In software-defined networking (SDN) programming, when a software-defined CPN access router cannot process data packets according to its local flow table, it reports the packet to the controller. The controller listens for these packet_in messages to implement the software-defined network functionality. For example... Figure 4 As shown, in this scheme, when the controller determines that the destination IP address of the ARP request is a service anycast address, it generates an ARP Reply packet based on the MAC address corresponding to the service anycast address and requests the access router to forward the response packet. When the controller determines that the message is a TCP / UDP request and the request destination is a service anycast IP, it first records the CPN access router that received the message, creates a service route entry based on the service subscription retrieval table, then generates a service forwarding policy based on the forwarding policy calculation algorithm, executes the service anycast forwarding queue update process based on the service forwarding policy to generate a service forwarding queue, and executes the following two computing power service forwarding flow table entries distribution:

[0092] 1. Item 1: Replace the destination address with the service instance address at the head of the forwarding queue, issue this item and request the switch to continue forwarding the data packets received by this access router;

[0093] 2. Item 2: If the destination address of the request matches the host address directly connected to this CPN access router, and the source address matches the maintenance service instance address of this access router, then the source address will be replaced with the service anycast address and this item will be issued.

[0094] Step S600: Determine the number of cloud nodes and edge nodes in the queuing queue based on the weight values ​​of cloud nodes and edge nodes, construct the queuing queue, and distribute the processing tasks corresponding to the service requests to the corresponding cloud nodes or edge nodes based on the queuing queue.

[0095] In practical implementation, this solution further divides the computing power network service anycast forwarding system architecture into data plane functions and control plane functions. The data plane functions include:

[0096] 1. Service Request Processing Based on Circular Forwarding Queues: In traditional target-based forwarding mechanisms, the selection of target nodes is typically based on static metrics such as latency and hop count. However, in the computing network designed in this invention, a dynamic circular forwarding queue mechanism is introduced into the data plane. When a service request arrives and matches a service ID, the queue will use the target computing node at the front of the queue as the forwarding target. This mechanism ensures that, under a well-designed request forwarding strategy, requests can be evenly distributed across multiple service replicas with an update-granular approach, preventing overload on certain nodes. After forwarding, the target is moved to the end of the queue to ensure the persistence of the strategy. The service anycast forwarding table is shown in Table 1. This table belongs to the data channel ID of the CPN access router and includes information such as the target matching service ID, forwarding queue, queue issuance time, queue expiration time, and the number of service ID matches.

[0097] Table 1

[0098]

[0099] Service-based encapsulation and decapsulation: In some complex network environments, the data plane is also responsible for encapsulating and decapsulating external headers for cross-domain service requests, ensuring that requests can be processed across different network regions or devices.

[0100] Control plane functions include:

[0101] 1. Information perception layer functions, specifically including:

[0102] a. Service Request Rate Prediction: By analyzing service request statistics reported by access routers, information sensing can predict the request arrival rate in real time over a future period. This helps to pre-schedule computing resources, optimize service response speed, and reduce the startup time of service replicas.

[0103] b. CPN Topology Awareness: The information awareness layer continuously monitors the CPN's network topology, including the location of compute nodes, the status of network links, and the computational load of each node. This awareness function provides real-time data support for calculating forwarding strategies.

[0104] c. Service Instance Discovery: The information awareness layer learns about the currently deployed service instances, including the number, location, and processing capacity of service replicas, through information reported by each egress router. This helps to dynamically adjust anycast forwarding strategies and achieve load balancing.

[0105] 2. Forwarding policy calculation layer functions, specifically including:

[0106] a. Service Subscription Retrieval: When a user request reaches the access router, if the router does not have a matching forwarding entry for the service request, it reports the request to the control plane. The control plane first performs a service subscription retrieval, searching for suitable service instances that match the user request. As shown in Table 2, the control plane service subscription retrieval table for computing power networks, the control plane function also needs to record the CPN access router ID information in the list of subscribed service routers. The control plane service subscription retrieval table includes the service ID, a list of available service instances, the service instance's processing capacity per unit time, the information update time, the information expiration time, and the list of CPN access routers subscribed to this service.

[0107] Table 2

[0108]

[0109] b. QoS Identification and Forwarding Queue Generation: The forwarding policy calculation layer identifies the priority of requests based on the QoS requirements of the service, such as bandwidth, latency, jitter, etc., and generates anycast forwarding queues according to the forwarding policy calculation method to ensure that service requests can be forwarded to the target computing node in the most balanced way.

[0110] c. Forwarding strategy optimization: The computing layer is also responsible for continuously optimizing existing forwarding strategies, such as using AI algorithms for global path optimization to improve the utilization efficiency of network resources.

[0111] By adopting the above scheme, this scheme achieves an organic combination of centralized control and distributed operation through the separation of the control plane and data plane of software-defined networking (SDN). In the complex environment of cross-layer nodes, the computing power network service anycast forwarding architecture fully utilizes the global optimization capability of centralized control and the flexibility of the queue forwarding mechanism based on weighted strategy in the distributed data plane. This enables the system to perform fine-grained balanced scheduling of burst traffic and traffic fluctuations from the perspective of global optimization, while ensuring stable service quality under high load conditions and improving resource utilization.

[0112] like Figure 2 As shown, in the architecture of the computing power network service anycast forwarding system, the service request processing flow is as follows:

[0113] 1. Service Request Arrival: When a user sends a service request, the request first reaches the nearest CPN access router. The access router first checks its local forwarding table to see if there is already a forwarding path for the service. If a valid path for the service is already in the forwarding table, it is forwarded directly according to the forwarding queue.

[0114] 2. Service Subscription Request: The access router uses a publish-ordering model to record currently processing but incomplete requests. This mechanism reduces the search space for service requests in the network, thereby speeding up request processing. Specifically, if the access router cannot find a matching service forwarding path, it sends a service subscription request to the controller. This request contains key information such as the type of service required by the user and QoS requirements. The controller retrieves deployed service instances by subscribing to services and selects a set of service instances based on factors such as the service's geographical location, computing power, and current load, recording them in the controller's service subscription record list.

[0115] 3. Generate and distribute anycast forwarding queue: After receiving a service subscription request, the controller, combining information from the CPN's topology awareness and service instance discovery modules, generates an anycast forwarding queue for that user's service request. This queue contains multiple target compute nodes, and the frequency of different target compute nodes appearing in the forwarding queue is set according to the forwarding policy requirements to ensure balanced service quality. The controller then distributes this forwarding queue to the access routers.

[0116] 4. Forwarding Execution: After receiving the anycast forwarding queue from the controller, the access router forwards service requests according to the order in the queue. Simultaneously, if the request requires cross-domain transmission, the access router will encapsulate the request to ensure it can be transmitted between different networks.

[0117] 5. Response Return: After the service replica instance on the compute node has finished processing the request, the response data will be returned to the user along the original path.

[0118] In some embodiments of the present invention, the method includes the following steps: when the controller receives a service request for the first time, the controller parses the service type corresponding to the service request, calculates the network transmission cost matrix and arrival rate matrix corresponding to the service type, and calculates the queuing queue corresponding to the service type.

[0119] When the controller receives a service request again, it parses the service type corresponding to the service request and determines whether there is a queue for that service type.

[0120] If it exists, the processing task corresponding to the service request will be directly distributed to the corresponding cloud node or edge node based on the existing queue.

[0121] If it does not exist, then construct a queue corresponding to the service type.

[0122] In some embodiments of the present invention, in the step of sending the processing task corresponding to the service request to the corresponding cloud node or edge node based on the queue, the cloud node or edge node at the head of the queue is used as the sending cloud node or edge node, and the cloud node or edge node is moved to the tail of the queue to update the queue.

[0123] In some embodiments of the present invention, in the step of calculating the transmission cost and propagation cost from the gateway to each cloud node and edge node based on the transmission distance and bandwidth between the gateway and the cloud node and the edge node, and calculating the network transmission cost matrix based on the transmission cost and propagation cost:

[0124] Based on the pre-set data packets, the amount of input and output data transmitted from the gateway to each cloud node and edge node is used to calculate the transmission cost based on the input and output data amounts and bandwidth; the time consumed by the pre-set data packets from the gateway to each cloud node and edge node is used as the propagation cost.

[0125] Based on the transmission cost and propagation cost, the gateway transmits the cost to each cloud node and edge node, and constructs a network transmission cost matrix.

[0126] In some embodiments of the present invention, in the step of calculating the transmission cost based on the input data volume and output data volume transmitted from the gateway to each cloud node and edge node based on a preset data packet, and the bandwidth, the transmission cost is calculated based on the following formula:

[0127]

[0128] in, This represents the transmission cost value from router i in the gateway to the cloud node and edge node j; This represents the pre-set amount of input data for service k, transmitted from router i of the gateway to the cloud node and edge node j. This indicates the pre-configured amount of data packets for service k that are transmitted from router i of the gateway to the cloud node and edge node j. This indicates the bandwidth used to transmit pre-configured data packets from router i in the gateway to the cloud node and edge node j.

[0129] In some embodiments of the present invention, in the step of calculating the target anycast forwarding weight strategy matrix using the Sinkhorn algorithm based on the arrival rate matrix and the network transmission cost matrix, the arrival rate matrix and the network transmission cost matrix are used as two diagonal matrices of the Sinkhorn algorithm, and an objective function is constructed based on the arrival rate matrix and the network transmission cost matrix. The final objective function value is determined by iteratively performing the Sinkhorn algorithm to obtain the target anycast forwarding weight strategy matrix.

[0130] In its implementation, the Kantorovich relaxed transport problem considers how to map multiple data points existing in one space to multiple data points in another space. It does not assume large-scale transport should be deterministic, but rather probabilistic. This probabilistic transport allows for the large-scale splitting from one source to several destinations, meaning that quality at any given point may be distributed across several locations. In CPN networks, the service anycast forwarding problem has evolved from a traditional end-to-end, one-to-one routing problem to a planning problem that simultaneously considers service requests initiated by multiple nodes and forwards these requests to multiple different service sites. Given this model similarity, this invention uses Kantorovich relaxed transport to model the CPN service anycast forwarding problem. The basic approach of this scheme is to first obtain the processing capacity of the service instance, then constrain the request rate reaching the service instance as a constraint on the Sinkhorn algorithm to ensure that the computation time of the request is less than a certain threshold, and then optimize network latency.

[0131] In some embodiments of the present invention, in the step of constructing an objective function based on the arrival rate matrix and the network transmission cost matrix, the objective function is constructed as follows:

[0132]

[0133] in, Represents the objective function value. Let the arrival rate matrix represent the arrival rate for service k. This represents the network transmission cost matrix for service k. This indicates the calculation of the inner product of matrices. This indicates minimization.

[0134] In some embodiments of the present invention, in the step of calculating the arrival rate matrix based on the forwarding probability weights of the gateway, cloud node, and edge node, the arrival rate matrix is ​​calculated based on the following formula:

[0135]

[0136] in, Represents the arrival rate matrix. Represents the rate vector. This represents the forwarding strategy matrix.

[0137] In the specific implementation process, use Indicates the number of connected routers, using Indicates the number of cloud nodes and edge nodes used to deploy the service. Let the reach rate vector represent the rate at which user service requests reach the CPN access router i and request service k within a unit of time. Represented as:

[0138]

[0139] use The request processing capacity vector represents the request processing capacity of service instance k of service k per unit time. Represented as:

[0140]

[0141] Network transmission cost matrix :

[0142]

[0143] definition Indicates CPN access router Forward the request for service k to the computing site. The probability weights. Then the forwarding strategy matrix. The definition is as follows:

[0144]

[0145] For CPN access routers It is noted that it will forward and offload traffic from k to the computing site. The request delivery rate Then there is

[0146]

[0147] Kantorovich's relaxed transport model generally assumes that transport is balanced. For example, consider the example of raw materials being transported to a factory: the amount of raw materials needed by the "factory" equals the amount of raw materials stored in the warehouse. When these two are equal, the problem model is an equality-constrained linear programming model, which can be easily solved quickly using iterative algorithms. When the quantities are unequal, the transport problem becomes an unbalanced transport problem. In the unbalanced case:

[0148] 1. The total processing capacity of the distributed, deployed service instances exceeds the user's needs.

[0149] In this case, our method assumes the existence of a virtual CPN access router, and that the requests collected by this router just meet the unbalanced demand. Since this router node does not actually exist, and therefore no additional user demand exists, it should not affect the overall cost band function. Setting the transmission cost to 0 eliminates the impact of virtual requests on the optimization results.

[0150] 2. The total processing capacity of the distributed, deployed service instances is lower than the user's needs.

[0151] When this situation occurs, it indicates that the current distributed deployment scheme for service k is unreasonable, and the processing capacity or the number of service replicas should be increased. Unlike the warehousing and transportation problem, where the amount of raw materials transported can be reduced when the inventory exceeds the factory's demand, in a CPN, when a service request arrives, the routing node must promptly look up and forward it. Therefore, it is not possible to forcibly reduce the request arrival rate in the same way as reducing the amount of raw materials transported. Furthermore, the total processing capacity of the deployed application services is insufficient. In this case, a simple linear simulation can be used to increase the processing capacity of the application services to match the user demand. This approach essentially uses the processing capacity of each distributed service instance as a weight to bear the user demand that exceeds the overall system load capacity, defining a coefficient... : Adjust the processing capacity vector of service instances in a distributed deployment. The pseudocode for the service anycast strategy calculation algorithm based on optimal transmission is as follows: Figure 5 As shown.

[0152] In some embodiments of the present invention, in the step of calculating the network transmission cost matrix based on the transmission cost and propagation cost, the transmission cost value transmitted from the gateway router to each cloud node and edge node is calculated, and the propagation cost value transmitted from the gateway router to each cloud node and edge node is measured by a preset data packet for service k. For each cloud node or edge node, the sum of the transmission cost value and the propagation cost value is calculated as the value at each position in the network transmission cost matrix, and the network transmission cost matrix is ​​constructed.

[0153] The beneficial effects of this plan include:

[0154] 1. Computing Power Network Service Anycast Forwarding System Architecture. This invention proposes a service anycast forwarding architecture for computing power networks, including a central controller, a cluster of computing nodes, a network topology awareness module, a service quality monitoring module, and a weighted policy forwarding queue. This architecture, by separating the control plane and the data plane, enables dynamic anycasting of computing resources and service requests across multiple node layers, solving the problem that traditional network architectures struggle to meet dynamic computing service demands and supporting fine-grained load balancing for differentiated service instance processing capabilities.

[0155] 2. Implementation and Key Functional Design of the SDN Controller under the Service Anycast Forwarding Architecture. This invention uses the RyuOpenFlow controller platform to implement the control plane functions for service anycast forwarding in the computing network, including key functions such as service anycast forwarding entry classes, real-time load status updates, computing network service anycast forwarding queue update process, and computing network service control plane Packet_in message processing process. Through the centralized management function of the SDN controller, real-time monitoring and dynamic scheduling of cross-layer computing nodes are achieved, and a modular implementation process design is provided, enabling the computing network service anycast forwarding mechanism to have the ability to be quickly deployed and flexibly expanded.

[0156] 3. Service Anycast Strategy Model and Algorithm Based on Kantorovich Relaxed Transport. To address the issue of fluctuating computing resource load across different nodes in a computing network, a load-aware forwarding model based on Request Processing Capability (RPC) is designed. For the problem of dynamic forwarding optimization of service requests, this invention establishes an anycast forwarding model based on Kantorovich relaxed transport and designs a service anycast strategy calculation algorithm based on optimal transmission. This ensures that the differences in processing capabilities and network costs between service instances are effectively utilized during resource scheduling to optimize service request forwarding decisions, further improving the overall transmission efficiency and service quality of the network.

[0157] 4. Flexible Centralized-Distributed Control Architecture. This invention achieves an organic combination of centralized control and distributed operation by separating the control plane and data plane of Software-Defined Networking (SDN). In complex environments with cross-layer nodes, the computing power network service anycast forwarding architecture fully leverages the global optimization capabilities of centralized control and the flexibility of the weighted queue forwarding mechanism in the distributed data plane. This enables the system to perform fine-grained balanced scheduling of burst traffic and traffic fluctuations from a globally optimal perspective, while ensuring stable service quality under high load conditions and improving resource utilization.

[0158] 5. Efficient and Stable Anycast Strategy Calculation. This invention introduces the Kantorovich relaxed transport model and designs an optimal anycast algorithm for transmission services, balancing the relationship between computational latency and network latency. Based on weighted strategy-based forwarding queues and request processing capacity (RPC) metrics, transmission performance fluctuations caused by load information fluctuations are reduced. Thus, even under traffic fluctuations, relatively stable anycast service performance optimization can be achieved, thereby ensuring service quality.

[0159] 6. Cross-domain and multi-layer node collaborative scheduling capability. The computing power network service anycast forwarding system architecture of this invention possesses resource scheduling capabilities across multiple layers and domains, enabling efficient allocation and forwarding of service requests among nodes at different levels, such as cloud and edge. Employing a stream-based request forwarding mechanism and optimizing forwarding strategies based on queue theory, it achieves collaborative work among multi-layer nodes, significantly reducing end-to-end latency and meeting the diverse user needs for low-latency and high-performance services, thereby improving overall service quality and system stability.

[0160] This invention also provides a computing power network service anycast system. The system includes a computer device, which includes a processor and a memory. The memory stores computer instructions, and the processor executes the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps of the method described above.

[0161] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned computing power network service anycast method. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disks, removable storage disks, CD-ROMs, or any other form of storage medium known in the art.

[0162] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0163] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0164] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0165] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for anycasting computing power network services, characterized in that, This method is based on a pre-built computing power network, which includes a gateway, a controller, multiple cloud nodes, and multiple edge nodes. The steps of this method include: Based on the gateway receiving the service request, the gateway forwards the service request to the controller; The controller acquires the working status of each cloud node and edge node in real time, and constructs the cloud nodes and edge nodes into a node set; Based on the transmission distance and bandwidth between the gateway and the cloud nodes and edge nodes, calculate the transmission cost and propagation cost from the gateway to each cloud node and edge node, and calculate the network transmission cost matrix based on the transmission cost and propagation cost; The arrival rate matrix is ​​calculated based on the forwarding probability weights of the gateway, cloud nodes, and edge nodes; Based on the arrival rate matrix and the network transmission cost matrix, the Sinkhorn algorithm is used to calculate the target anycast forwarding weight strategy matrix, which includes the weight values ​​of each cloud node and edge node. The number of cloud nodes and edge nodes in the queuing queue is determined based on the weight values ​​of cloud nodes and edge nodes, and the queuing queue is constructed. Based on the queuing queue, the processing tasks corresponding to the service requests are distributed to the corresponding cloud nodes or edge nodes.

2. The computing power network service anycast method according to claim 1, characterized in that, The method includes the following steps: when the controller receives a service request for the first time, the controller parses the service type corresponding to the service request, calculates the network transmission cost matrix and arrival rate matrix corresponding to the service type, and calculates the queuing queue corresponding to the service type. When the controller receives a service request again, it parses the service type corresponding to the service request and determines whether there is a queue for that service type. If it exists, the processing task corresponding to the service request will be directly distributed to the corresponding cloud node or edge node based on the existing queue. If it does not exist, then construct a queue corresponding to the service type.

3. The computing power network service anycast method according to claim 1 or 2, characterized in that, In the step of sending the processing task corresponding to the service request to the corresponding cloud node or edge node based on the queue, the cloud node or edge node at the head of the queue is used as the sending cloud node or edge node, and the cloud node or edge node is moved to the tail of the queue to update the queue.

4. The computing power network service anycast method according to claim 1, characterized in that, In the step of calculating the transmission cost and propagation cost from the gateway to each cloud node and edge node based on the transmission distance and bandwidth between the gateway and the cloud node and the edge node, and calculating the network transmission cost matrix based on the transmission cost and propagation cost: Based on the pre-set data packets, the amount of input and output data transmitted from the gateway to each cloud node and edge node is used to calculate the transmission cost based on the input and output data amounts and bandwidth; the time consumed by the pre-set data packets from the gateway to each cloud node and edge node is used as the propagation cost. Based on the transmission cost and propagation cost, the gateway transmits the cost to each cloud node and edge node, and constructs a network transmission cost matrix.

5. The computing power network service anycast method according to claim 4, characterized in that, In the step of calculating the transmission cost based on the input and output data volumes transmitted from the gateway to each cloud node and edge node according to pre-set data packets, and the bandwidth, the transmission cost is calculated based on the following formula: in, This represents the transmission cost value from router i in the gateway to the cloud node and edge node j; This represents the pre-set amount of input data for service k, transmitted from router i of the gateway to the cloud node and edge node j. This indicates the pre-configured amount of data packets for service k that are transmitted from router i of the gateway to the cloud node and edge node j. This indicates the bandwidth by which the pre-configured data packets are transmitted from router i of the gateway to the cloud node and edge node j.

6. The computing power network service anycast method according to claim 1, characterized in that, In the step of calculating the target anycast forwarding weight strategy matrix using the Sinkhorn algorithm based on the arrival rate matrix and the network transmission cost matrix, the arrival rate matrix and the network transmission cost matrix are used as the two diagonal matrices of the Sinkhorn algorithm, and an objective function is constructed based on the arrival rate matrix and the network transmission cost matrix. The final objective function value is determined by iteratively using the Sinkhorn algorithm, and the target anycast forwarding weight strategy matrix is ​​obtained.

7. The computing power network service anycast method according to claim 6, characterized in that, In the step of constructing the objective function based on the arrival rate matrix and the network transmission cost matrix, the objective function is constructed as follows: in, Represents the objective function value. Let the arrival rate matrix represent the arrival rate for service k. This represents the network transmission cost matrix for service k. This indicates the calculation of the inner product of matrices. This indicates minimization.

8. The computing power network service anycast method according to claim 7, characterized in that, In the step of calculating the arrival rate matrix based on the forwarding probability weights of the gateway, cloud node, and edge node, the arrival rate matrix is ​​calculated based on the following formula: in, Represents the arrival rate matrix. Represents the rate vector. This represents the forwarding strategy matrix.

9. The computing power network service anycast method according to any one of claims 4 to 8, characterized in that, In the step of calculating the network transmission cost matrix based on the transmission cost and propagation cost, the transmission cost value from the gateway router to each cloud node and edge node is calculated, and the propagation cost value from the gateway router to each cloud node and edge node is measured by a pre-set data packet for service k. For each cloud node or edge node, the sum of the transmission cost value and the propagation cost value is calculated as the value at each position in the network transmission cost matrix, and the network transmission cost matrix is ​​constructed.

10. A computing power network service broadcasting system, characterized in that, The system includes a computer device, which includes a processor and a memory. The memory stores computer instructions, and the processor executes the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps of the method as described in any one of claims 1 to 9.

Images