Routing planning, device, equipment, storage medium and computer program product

Abstract

The invention discloses a routing planning method and device for a computing power network, equipment, a storage medium and a computer program product, and aims to solve the problem that the utilization rate of computing resources is relatively low due to congestion of computing nodes when an existing routing scheduling method is adopted. The method comprises the following steps: acquiring a plurality of computing tasks of a plurality of user services in a computing power network; according to the computing power node information of the computing power network and the computing task, respectively determining a computing node and an access node corresponding to the computing task; constructing a computing power network topology model according to the computing nodes and the access nodes; performing path solving on the computing power network topology model by utilizing a genetic algorithm to obtain a routing allocation result meeting the plurality of computing tasks; and according to the route distribution result, performing route planning on the plurality of calculation tasks of the plurality of user services.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a routing planning method, apparatus, device, storage medium and computer program product for a computing network. Background Art

[0002] With the rapid development of artificial intelligence, the demand for data, computing power and other resources for user tasks has increased significantly. The transmission, analysis and storage of massive data poses a huge challenge to traditional networks and cloud computing, which puts cloud computing and networks under greater pressure, driving computing to move from the cloud to the edge side close to the data source, forming decentralized computing resources in the network. Under this trend, the resources of decentralized single computing nodes are limited and it is difficult to meet user needs.

[0003] Therefore, while the network realizes the interconnection of distributed nodes, it also needs to have the ability to coordinate the scheduling of network and computing resources, and dynamically schedule tasks to the optimal computing nodes for processing through the optimal path. This future network architecture is named "computing network".

[0004] However, in traditional routing tasks, the calculation path planning (OSPF) is usually implemented based on the network delay, or the path planning (TE) is implemented based on traffic balancing. Since traditional routing tasks do not systematically analyze network jitter, packet loss and other information, the use of traditional path routing in the computing network will cause congestion of computing nodes and reduce the utilization of computing resources.

[0005] It can be seen from this that when the computing power network architecture realizes the interconnection of distributed nodes, how to achieve coordinated scheduling of computing power nodes through routing path planning so that the business can be scheduled to the optimal computing power equipment for processing has become a technical problem that needs to be solved urgently. Summary of the invention

[0006] The embodiment of the present application provides a routing planning method for a computing power network, which is used to solve the problem that the existing routing scheduling method will cause congestion of computing nodes and thus lead to low computing resource utilization.

[0007] The embodiment of the present application also provides a routing planning device for a computing power network, which is used to solve the problem that the existing routing scheduling method will cause congestion of computing nodes and thus lead to low computing resource utilization.

[0008] The embodiment of the present application also provides a routing planning device for a computing network, which is used to solve the problem that the existing routing scheduling method will cause congestion of computing nodes, thereby leading to low computing resource utilization.

[0009] The embodiment of the present application also provides a computer-readable storage medium to solve the problem that the existing routing scheduling method will cause congestion of computing nodes, thereby leading to low computing resource utilization.

[0010] The embodiment of the present application also provides a computer program product to solve the problem that the existing routing scheduling method will cause congestion of computing nodes, thereby leading to low computing resource utilization.

[0011] The present application embodiment adopts the following technical solutions: A method for routing planning of a computing power network, comprising: obtaining multiple computing tasks of multiple user services in the computing power network; determining computing nodes and access nodes corresponding to the computing tasks respectively according to computing power node information of the computing power network and the computing tasks; constructing a computing power network topology model according to the computing nodes and the access nodes; performing path solving on the computing power network topology model using a genetic algorithm to obtain routing allocation results that satisfy the multiple computing tasks; and performing routing planning for the multiple computing tasks of the multiple user services according to the routing allocation results.

[0012] A routing planning device for a computing network comprises: a data acquisition unit for acquiring network feature data to be analyzed; an autoencoder processing unit for inputting the network feature data to be analyzed into a pre-trained autoencoder network, and performing attribute reduction on the network feature data to be analyzed through the autoencoder network to obtain a feature category data set and a fault category node set; a dependency graph generation unit for generating a fault category dependency graph based on the feature category data set and the fault category node set, wherein the directed edges of the fault category dependency graph are used to represent the dependency relationship between each feature category in the feature category data set, and the conditional probability table CPT of each node of the fault category dependency graph is used to represent the degree of dependency between each feature category; and a fault root cause analysis unit for calculating the contribution of each feature category in the feature category data set to the network fault according to the fault category dependency graph, and determining the network fault category according to the contribution.

[0013] A routing planning device for a computing network, comprising: A processor; and a memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to perform the following operations: obtain multiple computing tasks of multiple user services in a computing power network; determine the computing nodes and access nodes corresponding to the computing tasks respectively according to the computing power node information of the computing power network and the computing tasks; construct a computing power network topology model according to the computing nodes and the access nodes; use a genetic algorithm to solve the path of the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks; and perform routing planning for the multiple computing tasks of the multiple user services according to the routing allocation result.

[0014] A computer-readable storage medium stores one or more programs, which, when executed by an electronic device including multiple application programs, enable the electronic device to perform the following operations: obtain multiple computing tasks of multiple user services in a computing power network; determine the computing nodes and access nodes corresponding to the computing tasks respectively according to the computing power node information of the computing power network and the computing tasks; construct a computing power network topology model according to the computing nodes and the access nodes; use a genetic algorithm to solve the path of the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks; and perform routing planning for the multiple computing tasks of the multiple user services according to the routing allocation result.

[0015] A computer program product includes a computer program, which, when executed by a processor, implements: obtaining multiple computing tasks of multiple user services in a computing power network; determining computing nodes and access nodes corresponding to the computing tasks respectively according to computing power node information of the computing power network and the computing tasks; constructing a computing power network topology model according to the computing nodes and the access nodes; performing path solving on the computing power network topology model using a genetic algorithm to obtain routing allocation results that satisfy the multiple computing tasks; and performing routing planning for the multiple computing tasks of the multiple user services according to the routing allocation results.

[0016] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: By adopting the routing planning method of the computing power network provided in the embodiment of the present application, when performing routing planning of the computing power network, multiple computing tasks of multiple user services in the computing power network can be obtained, and the computing nodes and access nodes corresponding to each computing task can be determined respectively according to the computing power node information of the computing power network and the multiple computing tasks obtained, and according to the computing nodes and access nodes, the multi-access and multi-target computing power nodes are connected using virtual starting nodes and virtual target nodes to construct a computing power network topology model, thereby simplifying the problem into a single-to-single path planning problem from the virtual starting node to the virtual target node, and using a genetic algorithm to solve the path of the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks, and according to the routing allocation result, routing planning is performed for the multiple computing tasks of the multiple user services. By adopting the method provided in the embodiment of the present application, when performing route planning for a computing power network, first, the topological modeling of the computing power network is improved, and the problem of multiple access points and multiple available computing power nodes in the computing power network is simplified into a single-start, single-target path planning problem by combining virtual nodes. Compared with the existing method of separating the computing power nodes from the network routes and querying the optimal nodes in a polling manner, the computing efficiency is higher, the routing scheduling efficiency is greatly improved, and the congestion of computing nodes can be effectively avoided, thereby improving the utilization of computing resources; secondly, in the embodiment of the present application, the unsupervised learning method of genetic evolution algorithm is adopted to solve the path, thereby avoiding the problem that historical paths in the existing network are difficult to obtain and non-optimal, and compared with the supervised learning algorithm based on deep learning, the computing power required is small and no training is required. The optimal path planning can be completed using real-time data, which is more feasible and explainable in the routing planning task of the computing power network than the deep learning method. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A schematic diagram of a specific process of a routing planning method for a computing network provided in an embodiment of the present application; Figure 2 A specific flow chart of a path solving algorithm based on NSGA-II provided in an embodiment of the present application; Figure 3 A schematic diagram of the specific structure of a routing planning device for a computing network provided in an embodiment of the present application; Figure 4 A method provided in the embodiment of the present application Figure 2 A schematic diagram of the specific structure of the device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0019] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0020] The embodiment of the present application provides a routing planning method for a computing network, which is used to solve the problem that the existing routing scheduling method will cause congestion of computing nodes, thereby leading to low computing resource utilization.

[0021] The executor of the routing planning of the computing network provided in the embodiment of the present application may be, but is not limited to, at least one of a website server, a cloud platform server, a routing scheduling server, and a computing network server; in addition, the executor of the method may also be an application (Application, APP) running on these servers or the system itself.

[0022] For ease of description, the following takes the method being performed by a routing planning system as an example to introduce the implementation of the method. It is understandable that the method being performed by a routing planning system is only an exemplary description and should not be understood as a limitation on the method.

[0023] The specific implementation flow diagram of the routing planning method of the computing network provided in this application is as follows Figure 1 As shown, it mainly includes the following steps: Step 11, obtaining multiple computing tasks of multiple user services in the computing network; In an embodiment of the present application, there may be multiple users in the computing power network, and each user may correspond to one or more computing tasks (or computing services). In one implementation, in addition to obtaining the computing tasks sent by each user accessing the computing power network, the routing planning system also needs to obtain the task requirement parameters corresponding to each computing task.

[0024] In an embodiment of the present application, the task requirement parameters obtained by the routing planning system may specifically include: the geographic location information of the user corresponding to the computing task, the storage resource value required for the computing task, the computing resource value required for the computing task, the maximum business processing delay allowed by the computing task, the access delay from the computing task to the routing node, the size of the computing task, etc.

[0025] It should be noted here that the embodiments of the present application do not specifically limit which task requirement parameters are obtained and the method for obtaining these task requirement parameters.

[0026] Step 12, according to the computing tasks and computing node information of the computing network obtained by executing step 11, respectively determine the computing nodes and access nodes corresponding to each computing task; In an embodiment of the present application, whenever a computing task is obtained, the routing planning system can match available computing nodes for the computing task from the current computing power network based on the task requirement parameters corresponding to the computing task. At the same time, the routing planning system can also match available access nodes for the computing task from the current computing power network based on the geographic location information of the user corresponding to the computing task.

[0027] Step 13, constructing a computing power network topology model according to the available computing nodes and available access nodes determined by executing step 12; By executing the above step 12, the routing planning system will obtain multiple available computing nodes and multiple available access nodes corresponding to the computing task. In order to simplify the path planning problem of multiple access nodes and multiple computing nodes, in an embodiment of the present application, virtual nodes can be used to connect each computing node and access node respectively, and each computing node is connected to a virtual node, and the virtual node is used as a virtual target node, and each access node is also connected to another virtual node, and the virtual node is used as a virtual starting node, and then based on the virtual starting node (virtual Souce, VS) and the virtual target node (virtual Destination, VD), a computing power network topology model is constructed, thereby realizing the simplification of the path planning problem of multiple access nodes and multiple available computing nodes by using virtual nodes.

[0028] In the embodiment of the present application, the computing power network topology model can be constructed according to the following sub-steps, including: Sub-step 1301, connecting to an access node using a virtual start node; Sub-step 1302, connecting the virtual target node to the computing node; Sub-step 1303, connecting the access node with the computing power node according to the virtual start node and the virtual target node, and constructing a computing power network topology model; In one embodiment, the routing planning system can respectively obtain the network link status data and the computing power node status data in the computing power network, and respectively construct a link status model and a computing power status model based on the obtained network link status data and computing power node status data, and then construct constraint items based on whether the task requirements corresponding to the computing task match the network link status data and the computing power node status data in the computing power network. Finally, a computing power network topology model is constructed based on the link status model, the computing power status model and the constraint items.

[0029] Specifically, in the embodiment of the present application, the dynamic computing network involved in the embodiment of the present application can be defined according to the following formula [1]: [1] Where t represents the change of the computing network topology over time, V = {v1, v2, v3, ..., vn} is the set of nodes in the computing network topology, and E = {e1, e2, ..., e m} is the set of links in the computing network topology. Each link can be represented by two nodes at both ends, i.e., e i =(u, v), where u and v are both nodes in the node set V. The default network topology will not change in a short time. V is the set of state variables of the computing power node, and W E It is a collection of network link states. The states of links and nodes change over time.

[0030] In the embodiment of the present application, the requirements of the computing task for the network link status and computing node status can be determined based on the task requirement data obtained in the computing task. In one embodiment, the network link status data that the computing task generally needs to consider may include but is not limited to the following: link delay, jitter, packet loss rate, link traffic; and the information of the computing node status data that needs to be considered considers the computing cost, available memory, available storage, and computing time overhead.

[0031] Specifically, in the process of constructing the computing network topology model, the embodiment of the present application selects multiple variables of network links and computing nodes in the computing network to construct a more realistic and comprehensive network topology. At the same time, combined with the business needs of the computing task scheduling scenario, the constraints and optimization goals of these variables are defined as follows: Constraints: The constraints of the optimization model are whether the task requirements match the current network link status and computing node status, which can be specifically divided into: whether the link bandwidth matches, whether the available storage matches, and whether the available memory matches. In one implementation, the specific constraints may include: The task required bandwidth must be less than the available bandwidth of the current network link, denoted as c1.

[0032] The task requires computing storage that is larger than the currently available storage capacity, denoted as c2.

[0033] The task requires that the computing memory be smaller than the currently available content capacity, denoted as c3.

[0034] Optimization goal: The optimization model includes two attributes: computing power and network, which can be divided into the following five different goals: 1. Link delay is the sum of the delays of each link in the routing path, and the optimization goal f1 is to minimize the link delay; 2. Link jitter is the maximum jitter of each link in the routing path, and the optimization goal f2 is to minimize the link jitter; Packet loss rate is the maximum packet loss of each link in the routing path. The optimization goal f3 is to minimize the packet loss rate; Calculate costs The computing cost can be obtained by the total computing power required by the computing task * the unit computing power cost of the computing node. The optimization goal f4 is to minimize the computing cost; Computation time overhead The computing time overhead is the quotient of the computing amount required for the computing task and the computing power per second that the computing power node can provide. The optimization goal f5 is to minimize the computing time overhead.

[0035] Therefore, the model optimization problem can be defined according to the following formula [2]: [2] In an embodiment of the present application, a key to using a genetic algorithm to solve the path of the above-mentioned computing power network topology model is to construct a mapping relationship between routing links and genotypes. In order to enable any genotype to decode a complete routing path from the access node to the target computing node during the genetic evolution process, thereby performing fitness calculation, a coding and decoding strategy based on node priority is designed in an embodiment of the present application. The routing planning system can construct a routing path to be selected based on a virtual starting node and a virtual target node, and then encode the routing path to be selected according to a pre-set coding algorithm to obtain a coded routing path.

[0036] During the encoding process, the value of each gene in the chromosome is a non-negative integer, and any two genes have different values, representing the weights of the nodes contained in the path. In order to ensure the legitimacy of the path during the decoding process, an end-to-end decoding algorithm based on node priority genotype is designed.

[0037] In the node priority genotype end-to-end decoding algorithm, the output path and the current node c can be initialized based on the virtual starting node VS, and the restricted setting of the virtual starting node VS is set to 0, indicating that the node has been visited to prevent the formation of a loop. The algorithm searches for the next hop node in a loop until the next hop node belongs to the destination node set. In each cycle, the node with the highest weight is selected from the neighboring nodes as the next hop node; if the node l is visited for the first time, it is included in the path and the weight of the visited node is set to 0. If the node l is revisited, the algorithm will return to the previous node because the weights of all nodes adjacent to the previous node are also 0; if the destination node is still not found after traversing all nodes, the algorithm returns to the source node and returns an empty set, indicating that there is no available computing power node for the current access task.

[0038] Step 14, using a genetic algorithm to perform path solving on the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks; In the process of routing path planning of computing power network, it involves constraints and optimization objectives of multiple states including network links and computing nodes, which is a typical multi-objective path planning problem. In order to achieve fast and accurate solution, in the embodiment of the present application, the routing planning system can solve the path based on the multi-population genetic algorithm improved by NSGA-II to obtain the routing allocation results that meet multiple computing tasks.

[0039] In the embodiment of the present application, the overall block diagram of the solution algorithm is as follows: Figure 2 As shown in the figure. In this algorithm, all populations are first randomly initialized, and then different control parameters are assigned to each population. Each population evolves separately, and the individuals in each population are connected and migrated through immigration operations. The best individuals in each generation of evolution of each population form an elite population through the elite retention strategy. The population size of different sub-populations is dynamically adjusted through dynamic adjustment strategies, and individuals in the elite population will be added to each population with excellent performance. The elite population will also serve as the basis for the algorithm to end iteration and converge.

[0040] In the embodiments of the present application, the following improvements are made on multi-objective optimization based on NSGA-II on the basis of the traditional genetic algorithm.

[0041] A. Improved congestion calculation The selection mechanism of the genetic algorithm makes it easy for the solution to automatically converge to a certain area during the iteration process. In actual business, it is often hoped to ensure the diversity of solutions. In order to prevent the algorithm from falling into the local optimum too early, the concept of congestion distance is introduced in the NSGA-II algorithm in the embodiment of the present application to measure the degree of difference between solutions. The congestion distance of two solutions P and Q is defined as the sum of the distance difference between the two individuals on each sub-objective function, that is, the sum of the length and width of the rectangle.

[0042] B. Fast non-dominated sort An important design of NSGA-II is the fast non-dominated sorting algorithm, which significantly reduces the computational complexity. u is the number of dominant individuals x in the current population, A x is the set of individuals in the population that are dominated by individual x. In the embodiment of the present application, the following sub-steps can be used to perform non-dominated sorting on the individuals in the population, including: Sub-step b-1, let h = 1, for each individual x, if its X is determined and calculated through the dominance relationship u =0, put it into the set Rank h middle; Sub-step b-2, delete Rank in the population h , let h = h + 1, and recalculate X for each individual x in the remaining population u value, X u = 0 and put it into the set Rank h , as the next level of non-dominated solution set; Sub-step b-3, repeat sub-step b-2 until all individuals are stratified. After stratification, each individual in the lower Rank set has dominance over all individuals in the higher Rank set.

[0043] C. Elite retention strategy In the embodiment of the present application, the selection of the multi-objective genetic algorithm is implemented by an elite retention strategy, and its specific algorithm steps are shown in the following sub-steps: Sub-step c-1, after the crossover and mutation operations, in order to fully ensure the diversity of the population, the U offspring individuals generated are merged with the parent population to obtain a population with 2U individuals, and the fitness of each individual indicator in the population is calculated; Sub-step c-2, in the fast non-dominated sort, let the set of dominated solutions eliminated for the xth time be Rank x , where the number of individuals is NUM (Rank x). Because individuals in lower Ranks have dominance over all individuals in higher Ranks (i.e., better solutions), all individuals in lower Ranks 1, Rank 2, Rank 3, etc. are selected in turn until a certain Rank j After all individuals are selected, the number of individuals in the next generation population exceeds U, as shown in the following formulas [3] and [4]: [3] [4] At the same time, in order to ensure that the total number of individuals in each generation of the population remains unchanged, some individuals of the jth level Rank should be selected according to the crowding order so that the total number of individuals retained in the offspring is U.

[0044] Sub-step c-3, according to the crowding distance from large to small, put some individuals in the j-th level Rank into the next generation population. The elite strategy of sorting the 2U individuals of the offspring and the parent generation together can ensure that the best individuals of the population will not disappear due to genetic operations such as hybridization and mutation, thereby ensuring the convergence of the genetic algorithm.

[0045] Since the initialization process and evolution process of the genetic algorithm are both random, it cannot guarantee that the global optimal solution can be found, that is, it cannot guarantee that the optimal computing node and routing path can be found. The use of multiple populations to evolve independently can significantly reduce the possibility of falling into the local optimum. At the same time, the multi-population optimization strategy realizes the connection between populations by setting up immigration operations, replacing the poor individuals in the target population with the better individuals in the original population. The selected better individuals will participate in the subsequent iterative evolution process of the target population. Each population has its own control parameters and can search in different optimization directions in the solution space, thereby achieving the goal of multi-population collaborative optimization.

[0046] In addition, it should be noted that in an embodiment of the present application, the routing planning system can dynamically adjust the strategy to sort each subpopulation according to the quality of the best individual in each subpopulation, transfer the high-quality individuals in the subpopulation with a lower ranking to the population with a higher ranking, and dynamically adjust the search capability of each subpopulation by deleting a certain number of individuals in the population with a lower ranking, so as to achieve the purpose of accelerating the convergence speed.

[0047] Assume there are M populations, and the size of the mth population is Um. Set α (0 < α < 0.5) as the population adjustment parameter and β (0 < β < 0.5) as the individual adjustment parameter. Sort each subpopulation from the worst to the best according to the best individual quality. Select the first αM populations, and adjust the first βU m Individuals migrate to the M-α+1th subpopulation, and then delete the posterior βU in the αth subpopulation.m When it is detected that the number of individuals in the αth sub-population is less than or equal to 3, the dynamic adjustment of the sub-population is stopped.

[0048] By adopting the above steps, the computing power routing problem is abstracted into a multi-objective path optimization problem by modeling the dynamic topology of the computing power network, and the network indicators and computing power node indicators are optimized simultaneously. Then, through the priority-based encoding and decoding strategy, the network topology is converted into a genotype that can be solved by the genetic algorithm. Finally, combined with the NSGA based on multi-population improvement, the multi-objective path planning problem is solved by the genetic algorithm to obtain the optimal path solution set of the multi-state Pareto frontier.

[0049] Since each node in the computing network can provide computing resources, the solution set includes the optimal computing nodes and network path solutions under different attributes. It can be based on different business needs. For example, cloud gaming business needs to ensure the best network performance such as network latency, jitter, etc., and at the same time requires strong computing performance. At this time, higher weights are given to node computing power and network latency, so as to screen out the best computing routing planning path.

[0050] Step 15: Perform route planning for multiple computing tasks of multiple user services according to the route allocation result obtained by executing step 14.

[0051] By adopting the routing planning method of the computing power network provided in the embodiment of the present application, when performing routing planning of the computing power network, multiple computing tasks of multiple user services in the computing power network can be obtained, and the computing nodes and access nodes corresponding to each computing task can be determined respectively according to the computing power node information of the computing power network and the multiple computing tasks obtained, and according to the computing nodes and access nodes, the multi-access and multi-target computing power nodes are connected using virtual starting nodes and virtual target nodes to construct a computing power network topology model, thereby simplifying the problem into a single-to-single path planning problem from the virtual starting node to the virtual target node, and using a genetic algorithm to solve the path of the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks, and according to the routing allocation result, routing planning is performed for the multiple computing tasks of the multiple user services. By adopting the method provided in the embodiment of the present application, when performing route planning for a computing power network, first, the topological modeling of the computing power network is improved, and the problem of multiple access points and multiple available computing power nodes in the computing power network is simplified into a single-start, single-target path planning problem by combining virtual nodes. Compared with the existing method of separating the computing power nodes from the network routes and querying the optimal nodes in a polling manner, the computing efficiency is higher, the routing scheduling efficiency is greatly improved, and the congestion of computing nodes can be effectively avoided, thereby improving the utilization of computing resources; secondly, in the embodiment of the present application, the unsupervised learning method of genetic evolution algorithm is adopted to solve the path, thereby avoiding the problem that historical paths in the existing network are difficult to obtain and non-optimal, and compared with the supervised learning algorithm based on deep learning, the computing power required is small and no training is required. The optimal path planning can be completed using real-time data, which is more feasible and explainable in the routing planning task of the computing power network than the deep learning method.

[0052] In one implementation, the present application also provides a routing planning device for a computing network to solve the problem that the existing routing scheduling method causes congestion of computing nodes, thereby resulting in low computing resource utilization. Figure 3 As shown, it includes: a task acquisition unit 31, a node determination unit 32, a model construction unit 33, a solution unit 34 and a route planning unit 35.

[0053] The task acquisition unit 31 is used to acquire multiple computing tasks of multiple user services in the computing network; A node determination unit 32, configured to determine the computing node and the access node corresponding to the computing task according to the computing node information of the computing network and the computing task; A model building unit 33, configured to build a computing power network topology model according to the computing nodes and the access nodes; A solving unit 34 is used to solve the path of the computing power network topology model by using a genetic algorithm to obtain a routing allocation result that satisfies the multiple computing tasks; The route planning unit 35 is used to perform route planning for the multiple computing tasks of the multiple user services according to the route allocation result.

[0054] In one embodiment, the model building unit 33 is specifically used to: use a virtual starting node to connect with the access node; use a virtual target node to connect with the computing node; according to the virtual starting node and the virtual target node, connect the access node with the computing node to build a computing network topology model.

[0055] In one embodiment, the model building unit 33 is specifically used to: respectively obtain the network link status data and the computing power node status data in the computing power network; respectively build a link status model and a computing power status model according to the network link status data and the computing power node status data; build constraint items according to whether the task requirements corresponding to the computing task match the network link status data and the computing power node status data in the computing power network; and build a computing power network topology model according to the link status model, the computing power status model and the constraint items.

[0056] In one embodiment, it also includes an encoding unit, which is specifically used to: construct a candidate routing path according to the virtual starting node and the virtual target node; and encode the candidate routing path according to a preset encoding algorithm to obtain an encoded routing path.

[0057] In one implementation, the solving unit 34 is specifically used to: perform path solving on the computing power network topology model based on the NSGA-II genetic algorithm to obtain routing allocation results that meet the multiple computing tasks.

[0058] In one embodiment, the solution unit 34 is specifically used to: perform random population initialization processing on each node in the computing power network topology model to obtain multiple initial populations; sort each of the initial populations according to a fast non-dominated sorting algorithm, complete the selection of the initial population, and obtain multiple second initial populations; perform population crossover and mutation processing on the multiple second initial populations to complete population iteration and obtain a global optimal solution; based on the global optimal solution, obtain a routing allocation result that satisfies the multiple computing tasks.

[0059] By using the routing planning device of the computing power network provided in the embodiment of the present application, when performing routing planning of the computing power network, multiple computing tasks of multiple user services in the computing power network can be obtained, and the computing nodes and access nodes corresponding to each computing task can be determined respectively according to the computing power node information of the computing power network and the multiple computing tasks obtained, and according to the computing nodes and access nodes, the multi-access and multi-target computing power nodes are connected using virtual starting nodes and virtual target nodes to construct a computing power network topology model, thereby simplifying the problem into a single-to-single path planning problem from the virtual starting node to the virtual target node, and using a genetic algorithm to solve the path of the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks, and according to the routing allocation result, routing planning is performed for the multiple computing tasks of the multiple user services. By using the device provided in the embodiment of the present application, when performing route planning for a computing power network, first, the topological modeling of the computing power network is improved, and the problem of multiple access points and multiple available computing power nodes in the computing power network is simplified into a single-start, single-target path planning problem by combining virtual nodes. Compared with the existing method of separating the computing power nodes from the network routes and querying the optimal nodes in a polling manner, the computing efficiency is higher, the routing scheduling efficiency is greatly improved, and the congestion of computing nodes can be effectively avoided, thereby improving the utilization rate of computing resources. Secondly, in the embodiment of the present application, the unsupervised learning method of genetic evolution algorithm is used to solve the path, thereby avoiding the problem that historical paths in the existing network are difficult to obtain and non-optimal. Compared with the supervised learning algorithm based on deep learning, the computing power required is small and no training is required. The optimal path planning can be completed using real-time data. Compared with the deep learning method, it is more feasible and explainable in the route planning task of the computing power network.

[0060] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 4 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.

[0061] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0062] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0063] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a routing planning device for the computing power network at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations: obtain multiple computing tasks of multiple user services in the computing power network; determine the computing nodes and access nodes corresponding to the computing tasks according to the computing power node information of the computing power network and the computing tasks; construct a computing power network topology model according to the computing nodes and the access nodes; use a genetic algorithm to solve the path of the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks; and perform routing planning for the multiple computing tasks of the multiple user services according to the routing allocation result.

[0064] The above application Figure 4The method performed by the routing planning electronic device of the computing network disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in a decoding processor. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0065] Of course, in addition to software implementation methods, the electronic device of the present application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0066] The embodiment of the present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by a portable electronic device including a plurality of application programs, enable the portable electronic device to execute Figure 1 The method of the embodiment shown is specifically used to perform the following operations: Acquire multiple computing tasks of multiple user services in a computing power network; determine the computing nodes and access nodes corresponding to the computing tasks respectively according to the computing power node information of the computing power network and the computing tasks; construct a computing power network topology model according to the computing nodes and the access nodes; use a genetic algorithm to perform path solving on the computing power network topology model to obtain routing allocation results that satisfy the multiple computing tasks; and perform routing planning for the multiple computing tasks of the multiple user services according to the routing allocation results.

[0067] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0068] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0069] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0070] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0071] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0072] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0073] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0074] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0075] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0076] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for routing planning of a computing network, characterized in that: include: Obtain multiple computing tasks for multiple user services in the computing network; According to the computing power node information of the computing power network and the computing task, respectively determine the computing node and the access node corresponding to the computing task; Constructing a computing power network topology model according to the computing nodes and the access nodes; Using a genetic algorithm to solve the path of the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks; According to the route allocation result, route planning is performed for the multiple computing tasks of the multiple user services.

2. The method according to claim 1, characterized in that The constructing of a computing power network topology model according to the computing nodes and the access nodes specifically includes: Using a virtual starting node to connect to the access node; Using a virtual target node to connect to the computing node; According to the virtual starting node and the virtual target node, the access node is connected to the computing power node to construct a computing power network topology model.

3. The method according to claim 2, characterized in that The step of connecting the access node with the computing power node according to the virtual starting node and the virtual target node to construct a computing power network topology model specifically includes: Respectively obtain network link status data and computing power node status data in the computing power network; According to the network link status data and the computing power node status data, a link status model and a computing power status model are constructed respectively; According to the task requirements corresponding to the computing tasks, and whether the network link status data and the computing node status data in the computing network match, a constraint item is constructed; A computing power network topology model is constructed according to the link state model, the computing power state model and the constraints.

4. The method according to claim 3, characterized in that After constructing the computing power network topology model according to the computing nodes and the access nodes, the method further includes: Constructing a routing path to be selected according to the virtual starting node and the virtual target node; The candidate routing path is encoded according to a preset encoding algorithm to obtain an encoded routing path.

5. The method according to claim 1, characterized in that The using of a genetic algorithm to solve the path of the computing power network topology model specifically includes: The computing power network topology model is path-solved based on the NSGA-II genetic algorithm to obtain routing allocation results that meet the multiple computing tasks.

6. The method according to claim 5, characterized in that The path solving of the computing power network topology model based on the NSGA-II genetic algorithm to obtain the routing allocation results that meet the multiple computing tasks specifically includes: Performing random population initialization processing on each node in the computing power network topology model to obtain multiple initial populations; According to a fast non-dominated sorting algorithm, each of the initial populations is sorted to complete the selection of the initial population, and a plurality of second initial populations are obtained; Performing population crossover and mutation processing on the multiple second initial populations, completing population iteration, and obtaining a global optimal solution; According to the global optimal solution, a routing allocation result satisfying the multiple computing tasks is obtained.

7. A routing planning device for a computing network, characterized in that: include: A task acquisition unit, used to acquire multiple computing tasks of multiple user services in the computing network; A node determination unit, configured to determine a computing node and an access node corresponding to the computing task, respectively, according to the computing node information of the computing network and the computing task; A model building unit, used to build a computing power network topology model according to the computing nodes and the access nodes; A solution unit, used for performing path solution on the computing power network topology model by using a genetic algorithm to obtain a routing allocation result that satisfies the multiple computing tasks; A route planning unit is used to perform route planning for multiple computing tasks of the multiple user services according to the route allocation result.

8. A routing planning device for a computing network, comprising: processor; a memory arranged to store computer executable instructions which, when executed, cause the processor to: Obtain multiple computing tasks for multiple user services in the computing network; According to the computing power node information of the computing power network and the computing task, respectively determine the computing node and the access node corresponding to the computing task; Constructing a computing power network topology model according to the computing nodes and the access nodes; Using a genetic algorithm to solve the path of the computing power network topology model to obtain a routing allocation result that satisfies the multiple computing tasks; According to the route allocation result, route planning is performed for the multiple computing tasks of the multiple user services.

9. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including multiple application programs, enables the electronic device to execute the routing planning method for a computing power network as described in any one of claims 1-6.

10. A computer program product, characterized in that It includes a computer program, which, when executed by a processor, implements the routing planning method for the computing power network as described in any one of claims 1 to 6.

Images