Computing Task Processing Method, Apparatus, Device, and Medium

Abstract

The present application discloses a computing task processing method, apparatus, device and medium, relating to the field of computer technologies, including: obtaining a data packet of a current computing task of a target terminal and identification information of the current computing task, and determining the computing type of the current computing task by using the identification information; determining the path distances between all computing power nodes and an access router, and screening out a target computing power node from all computing power nodes based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information and the path distances, and planning a target transmission path between the target computing power node and the access router by using the software-defined network method and the segment routing technology; transmitting the data packet to the target computing power node by using a bandwidth reservation mechanism, and the target computing power node processes the data packet of the current computing task by using a resource reservation mechanism. High-quality computing task processing and high-quality computing task transmission can be achieved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly to a method, device, equipment and medium for processing computing tasks. Background Art

[0002] With the large-scale deployment of edge computing and intelligent terminal devices, it has become more convenient and efficient for user terminals to access and use these massive distributed computing resources. However, how to make more efficient use of these computing resources has become an urgent problem to be solved. A single edge computing node has limited resources and is difficult to effectively and quickly process computing tasks. Especially for some computationally intensive tasks, it may lead to heavy load on the edge computing node and long processing time of computing tasks. Although computing resources tend to be ubiquitously deployed, there is a lack of an effective coordination mechanism between edge computing nodes and between edge computing nodes and cloud computing nodes, and the allocation and scheduling mechanism of computing tasks is not yet perfect, resulting in low utilization rate of computing resources. Therefore, in order to make more efficient use of the massive distributed computing resources at the network edge and promote the deep integration and collaborative processing of distributed edge computing and the network, the concept of Compute Power Networking (CPN) has been proposed. The Compute Power Networking aims to connect and integrate distributed computing nodes and conduct overall scheduling, and through the improvement of network architecture and protocols, achieve the optimization and efficient utilization of network and computing resources. Since the concept of Compute Power Networking was proposed, it has received extensive attention from the academic and industrial communities. New services such as intelligent manufacturing and intelligent driving not only require high-performance computing resources but also require the network to provide deterministic and reliable guarantees to meet the functional and performance requirements of new services. However, the current network cannot achieve end-to-end deterministic transmission between computing nodes and efficient scheduling of computing power. This urgently requires the research on Compute Power Networking technology based on determinism to build a connected heterogeneous ubiquitous computing power node and ensure a deterministic computing power network with low latency, high reliability, and optimal path for transmission, so as to meet the end-to-end high-quality and deterministic transmission and scheduling of computing tasks between computing power nodes.

[0003] Traditional computing power network scheduling methods only solve the problem of whether computing tasks can be processed or there are computing tasks to be processed, while the problem that the computing power network should solve in the future is how to perform high-quality computing and high-quality transmission. In the current prior art, the distribution and scheduling strategy of computing tasks is based on the network link resource scheme, but it cannot fully ensure that the computing power nodes determined for processing computing tasks and the transmission paths during the transmission of computing tasks are both optimal, thus resulting in the inability to achieve truly high-quality computing and high-quality transmission.

[0004] In summary, how to achieve high-quality computing task processing and high-quality computing task transmission is a problem to be solved in this field. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a calculation task processing method, device, equipment and medium, which can realize high-quality calculation task processing and high-quality calculation task transmission. The specific scheme is as follows:

[0006] In a first aspect, the present application discloses a calculation task processing method, which is applied to a deterministic computing power network system, and includes:

[0007] Establish a transmission connection between the target terminal and the access router in the deterministic computing power network system, and based on the transmission connection, obtain the data packet of the current calculation task of the target terminal and the identification information of the current calculation task, and use the identification information to determine the calculation type of the current calculation task;

[0008] Respectively determine the path distances between all computing power nodes and the access router, and based on the calculation type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information and the path distances, screen out the target computing power nodes from all computing power nodes, and use software-defined network mode and segment routing technology to plan the target transmission path between the target computing power node and the access router;

[0009] Based on the target transmission path, and use the bandwidth reservation mechanism to transmit the data packet to the target computing power node, so that the target computing power node uses the resource reservation mechanism to process the data packet of the current calculation task.

[0010] Optionally, before obtaining the data packet of the current calculation task of the target terminal and the identification information of the current calculation task based on the transmission connection, it further includes:

[0011] If the type of the current calculation task is computationally intensive, set the preset identification position of the current calculation task through the target terminal to obtain the identification information of the current calculation task corresponding to the computationally intensive type;

[0012] And / or, if the type of the current calculation task is time-sensitive, set the preset identification position of the current calculation task through the target terminal to obtain the identification information of the current calculation task corresponding to the time-sensitive type.

[0013] Optionally, before screening out the target computing power nodes from all computing power nodes based on the calculation type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information and the path distances, it further includes:

[0014] Obtain the current computing power resource status information including the current network link status information, the computing power node service carrier information, and the current load information of all the computing power nodes.

[0015] Optionally, the screening of the target computing power nodes from all the computing power nodes based on the computing type, the current network link status information, the current computing power resource status information of all the computing power nodes, the computing power node service carrier information, and the path distance includes:

[0016] Based on the computing type, the current network link status information, the current computing power resource status information of all the current computing power nodes, the computing power node service carrier information, and the path distance, screen out the current computing power nodes from all the current computing power nodes, and calculate the transmission time of the data packet from the access router to the current computing power nodes. Determine whether the current computing power nodes meet the delay condition of the attribute information in the data packet based on the transmission time of the current computing power nodes;

[0017] If not, remove the current computing power nodes from all the current computing power nodes to obtain the next all computing power nodes, update the next all computing power nodes to all the current computing power nodes, and jump back to the step of screening out the current computing power nodes from all the current computing power nodes based on the computing type, the current network link status information, the current computing power resource status information of all the current computing power nodes, the computing power node service carrier information, and the path distance until the current computing power nodes meet the delay condition, so as to determine the current computing power nodes as the target computing power nodes.

[0018] Optionally, the screening of the current computing power nodes from all the current computing power nodes based on the computing type, the current network link status information, the current computing power resource status information of all the current computing power nodes, the computing power node service carrier information, and the path distance includes:

[0019] Determine the weight coefficients and scoring values of the computing type, the current network link status information, the current computing power resource status information of all the current computing power nodes, the computing power node service carrier information, and the path distance;

[0020] Use the weight coefficients to perform weighted calculation on the scoring values to obtain the weighted calculation results of all the current computing power nodes, and screen out the current computing power nodes from the weighted calculation results in descending order.

[0021] Optionally, the determination of the weight coefficients of the computing type, the current network link status information, the current computing power resource status information of all the current computing power nodes, the computing power node service carrier information, and the path distance includes:

[0022] If the current network link status information meets the first preset condition, reduce the weight coefficient of the current network link status information to the first preset threshold, and determine the calculation type, the current computing power resource status information of all current computing power nodes, the computing power node service carrier information, and the weight coefficient of the path distance;

[0023] If the current computing task meets the second preset condition, increase the weight coefficient of the calculation type to the second preset threshold, and determine the current network link status information, the current computing power resource status information, the computing power node service carrier information, and the weight coefficient of the path distance.

[0024] Optionally, the planning of the target transmission path between the target computing power node and the access router by using the software-defined network method and the segment routing technology includes:

[0025] Plan the transmission path between the target computing power node and the access router to obtain a path library containing several such transmission paths, and screen out the current transmission path from the path library in ascending order, and judge whether the task flow of the current transmission path is less than the preset task flow threshold;

[0026] If it is less, determine that the current transmission path is the target transmission path; if it is not less, remove the current transmission path from the path library, and re-jump to the step of screening out the current transmission path from several such transmission paths in ascending order until the task flow of the current transmission path is less than the preset task flow threshold, so as to obtain the target transmission path.

[0027] In a second aspect, the present application discloses a computing task processing device, which is applied to a deterministic computing power network system, and includes:

[0028] A path distance determination module, configured to obtain the data packet of the current computing task, the identification information of the current computing task, and the attribute information of the current computing task, and use the identification information to determine the calculation type of the current computing task, and determine the path distance between all computing power nodes and the target terminal;

[0029] A target node screening module, configured to screen out a target computing power node that meets the preset computing power node computing resource reservation mechanism from all the computing power nodes based on the attribute information, the current network link status information, the current computing power resource status information of all the computing power nodes, the computing power node service carrier information, and the path distance;

[0030] A processing module, configured to plan a target transmission path between the target computing node and the target terminal, and transmit the data packet to the target computing node based on the target transmission path, so that the target computing node processes the current computing task based on the data packet.

[0031] In a third aspect, the present application discloses an electronic device, including:

[0032] A memory, configured to store a computer program;

[0033] A processor, configured to execute the computer program to implement the steps of the computing task processing method disclosed above.

[0034] In a fourth aspect, the present application discloses a computer-readable storage medium, configured to store a computer program; wherein, when the computer program is executed by a processor, the steps of the computing task processing method disclosed above are implemented.

[0035] It can be seen that this application establishes a transmission connection between the target terminal and the access router in the deterministic computing power network system, and obtains the data packet of the current computing task and the identification information of the current computing task of the target terminal based on the transmission connection, and determines the computing type of the current computing task by using the identification information; respectively determines the path distances between all computing power nodes and the access router, and based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distance, filters out the target computing power node from all the computing power nodes, and uses the software-defined network method and the segment routing technology to plan the target transmission path between the target computing power node and the access router; based on the target transmission path, and uses the bandwidth reservation mechanism to transmit the data packet to the target computing power node, so that the target computing power node uses the resource reservation mechanism to process the data packet of the current computing task. It can be seen that when determining the target computing power node in this application, it is necessary to consider the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distance, so that the selected target computing power node can meet the network link status, computing power resource status, service carrier, computing type, and path distance required by the current computing task, and perform resource reservation processing on the target computing power node, that is, the subsequent target computing power node uses the resource reservation mechanism to process the data packet of the current computing task, thereby ensuring that the target computing power node has sufficient computing resources to process the current computing task, that is, ensuring that the current computing task can be processed with high quality; when planning the target transmission path between the target computing power node and the access router, the software-defined network method and the segment routing technology are used, so that the target transmission path is the optimal among all transmission paths, and because the bandwidth reservation mechanism is used when transmitting the data packet of the current computing task, it is ensured that the target transmission path has sufficient bandwidth resources to transmit the data packet of the current computing task to the target computing power node, that is, the data packet of the current computing task can be transmitted to the target computing power node with high quality. Description of the Drawings

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0037] Figure 1 It is a flowchart of a method for processing a computing task disclosed in the present application;

[0038] Figure 2A flowchart of a specific computing task processing method disclosed in this application;

[0039] Figure 3 A flowchart of a specific computing task processing method disclosed in this application;

[0040] Figure 4 A schematic diagram of a specific computing task processing disclosed in this application;

[0041] Figure 5 A schematic diagram of the structure of a computing task processing device disclosed in this application;

[0042] Figure 6 A structural diagram of an electronic device disclosed in this application. Detailed implementation manners

[0043] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0044] Traditional computing power network scheduling methods only solve the problem of whether a computing task can be processed or there is a computing task to be processed, while the problems that the computing power network should solve in the future are how to perform high-quality computing and high-quality transmission; in the current prior art, the distribution and scheduling strategy of computing tasks is based on the network link resource scheme, but it cannot fully ensure that the computing power nodes determined for processing computing tasks and the transmission paths during the transmission of computing tasks are all optimal, thus resulting in the inability to achieve true high-quality computing and high-quality transmission.

[0045] Therefore, this application correspondingly provides a computing task processing solution, which can achieve high-quality computing task processing and high-quality computing task transmission.

[0046] See Figure 1 As shown, an embodiment of this application discloses a computing task processing method, which is applied to a deterministic computing power network system and includes:

[0047] Step S11: Establish a transmission connection between the target terminal and the access router in the deterministic computing power network system, and obtain the data packet of the current computing task of the target terminal and the identification information of the current computing task based on the transmission connection, and determine the computing type of the current computing task by using the identification information.

[0048] In this embodiment, before obtaining the data packet of the current computing task of the target terminal and the identification information of the current computing task based on the transmission connection, the following steps are further included: if the type of the current computing task is compute-intensive, the target terminal sets the preset identification position of the current computing task to obtain the identification information of the current computing task corresponding to the compute-intensive type; and / or, if the type of the current computing task is time-sensitive, the target terminal sets the preset identification position of the current computing task to obtain the identification information of the current computing task corresponding to the time-sensitive type. It can be understood that the target terminal can add identification processing to the current computing task, that is, set the preset identification position of the current computing task to obtain the identification information corresponding to the computing type. The computing type can be compute-intensive or time-sensitive, or both compute-intensive and time-sensitive. For example, autonomous driving computing tasks, AR (Augmented Reality) / VR (Virtual Reality) computing tasks. These tasks require a large amount of computing and also need to be processed within a short time. Then it further shows that such computing tasks require a high transmission speed and processing speed.

[0049] It should be noted that the deterministic computing power network system includes a preset data plane and a preset control plane. Among them, the data plane includes access routers, and the preset data plane establishes a transmission connection between the target terminal and the access routers; the access routers obtain the data packet of the current computing task of the target terminal and the identification information of the current computing task based on the transmission connection, and report the data packet and the identification information to the preset control plane.

[0050] Step S12: Determine the path distances between all computing power nodes and the access router respectively, and based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distances, screen out the target computing power nodes from all computing power nodes, and use software-defined network methods and segment routing technologies to plan the target transmission path between the target computing power nodes and the access router.

[0051] In this embodiment, before screening out the target computing nodes from all the computing nodes based on the computing type, the current network link status information, the current computing power resource status information of all the computing power nodes, the computing node service carrier information, and the path distance, it further includes: obtaining the current network link status information, the computing node service carrier information, and the current computing power resource status information including the current load information of all the computing power nodes. The preset control plane receives the data packets and identification information reported by the preset data plane, and the preset control plane also needs to obtain the current network link status information, the computing node service carrier information, and the current computing power resource status information including the current load information of all the computing power nodes. Among them, the current network link status information refers to the status of the direct connection network between each current computing node and the corresponding router, and includes information about the network type and all the routers adjacent to the current router in the network; the current computing power resource status information may include the current parallel computing resources, the current serial computing resources, and the current load information, such as the current load conditions of the CPU (Central Processing Unit) and GPU (Graphics Processing Unit); the computing node service carrier information is the service carrier information in each computing node, and the service carrier is, for example, a virtual machine or a container.

[0052] It can be understood that in this embodiment, it is necessary to use the computing type, the current network link status information, the current computing power resource status information, the computing node service carrier information, and the path distance to screen out reasonable target computing nodes from all the computing nodes. The screening method is, for example, weighted calculation. Taking the computing type, the current network link status information, the current computing power resource status information, the computing node service carrier information, and the path distance as parameters, based on the actual situation of the current computing task, weight coefficients are assigned to each parameter. Because the computing type, the current network link status information, the current computing power resource status information, the computing node service carrier information, and the path distance are considered, the target computing node with the optimal current network link status, the best current computing power resource status, the computing node service carrier matching the current computing task and computing type, and the shortest path distance is screened out as much as possible, so as to be able to use the target computing node to process the current computing task with high quality. Among them, if there is no corresponding service carrier C in the computing node A with a shorter path distance, and the transmission time between the screened target computing node B and the access router is longer, then the serverless technology can be used to start the corresponding service carrier C in the computer in the computing node A and use the service carrier C to process the current computing task.

[0053] In this embodiment, the target transmission path between the target computing node and the access router is planned by using the software-defined network method and the segment routing technology. It can be understood that there may be multiple transmission paths between the target computing node and the access router. In order to quickly transmit the data packets of the current computing task, the transmission path can be shortened as much as possible, that is, the shortest transmission path is used as the target transmission path.

[0054] Step S13: Based on the target transmission path, use the bandwidth reservation mechanism to transmit the data packet to the target computing node, so that the target computing node uses the resource reservation mechanism to process the data packet of the current computing task.

[0055] In this embodiment, bandwidth reservation is the process of pre-assigning the corresponding bandwidth to the deterministic computing power network system and the application program of the target computing node. The allocation priority of the current computing task can be determined in advance according to the importance and / or delay sensitivity of the current computing task. In this way, the available bandwidth is utilized to the maximum extent, that is, the bandwidth required to transmit the current computing task is first determined. When transmitting the current computing task, this bandwidth is used for transmission to ensure that the current computing task has sufficient bandwidth resources for transmission. Also, because the target transmission path is shorter among all transmission paths, the transmission time can be shortened, further improving the transmission quality.

[0056] It can be understood that in this embodiment, resource reservation means that after determining the target computing node, the computing resources required to process the current computing task are determined in the target computing node. When the target computing node receives the data packet, it uses this computing resource to process the current computing task to ensure that there are sufficient computing resources to process the current computing task, that is, to process the current computing task with high quality.

[0057] It can be seen that the present application establishes a transmission connection between a target terminal and an access router in the deterministic computing power network system, and based on the transmission connection, obtains the data packet of the current computing task of the target terminal and the identification information of the current computing task, and determines the computing type of the current computing task by using the identification information; respectively determines the path distances between all computing power nodes and the access router, and based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distances, screens out target computing power nodes from all computing power nodes, and uses software-defined network mode and segment routing technology to plan the target transmission path between the target computing power node and the access router; based on the target transmission path, and uses a bandwidth reservation mechanism to transmit the data packet to the target computing power node, so that the target computing power node processes the data packet of the current computing task by using a resource reservation mechanism. Thus, it can be seen that when determining the target computing power node in the present application, it is necessary to take into account the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distances, so that the selected target computing power node can meet the network link status, computing power resource status, service carrier, computing type, and path distances required by the current computing task, and performs resource reservation processing on the target computing power node, that is, the subsequent target computing power node processes the data packet of the current computing task by using the resource reservation mechanism, thereby ensuring that the target computing power node has sufficient computing resources to process the current computing task, that is, ensuring that the current computing task can be processed with high quality; when planning the target transmission path between the target computing power node and the access router, the software-defined network mode and segment routing technology are used, so that the target transmission path is the optimal among all transmission paths, and because a broadband reservation mechanism is used when transmitting the data packet of the current computing task, it is ensured that the target transmission path has sufficient broadband resources to transmit the data packet of the current computing task to the target computing power node, that is, the data packet of the current computing task can be transmitted to the target computing power node with high quality.

[0058] See Figure 2 As shown, an embodiment of the present application discloses a specific computing task processing method, which is applied to a deterministic computing power network system, and includes:

[0059] Step S21: Establish a transmission connection between a target terminal and an access router in the deterministic computing power network system, and based on the transmission connection, obtain the data packet of the current computing task of the target terminal and the identification information of the current computing task, and determine the computing type of the current computing task by using the identification information.

[0060] Step S22: Determine the path distances between all computing nodes and the access router respectively, and based on the computing type, the current network link status information, the current computing power resource status information of all current computing nodes, the computing node service carrier information, and the path distances, screen out the current computing nodes from all current computing nodes, calculate the transmission time for the data packet to be transmitted from the access router to the current computing nodes, and determine whether the current computing nodes meet the delay condition of the attribute information in the data packet based on the transmission time of the current computing nodes.

[0061] In this embodiment, the screening out of the current computing nodes from all current computing nodes based on the computing type, the current network link status information, the current computing power resource status information of all current computing nodes, the computing node service carrier information, and the path distances includes: determining the weight coefficients and score values of the computing type, the current network link status information, the current computing power resource status information of all current computing nodes, the computing node service carrier information, and the path distances; using the weight coefficients to perform weighted calculation on the score values to obtain the weighted calculation results of all current computing nodes, and screening out the current computing nodes from the weighted calculation results in descending order. It can be understood that the sum of the weight coefficients of the parameters such as the computing type, the current network link status information, the current computing power resource status information, the computing node service carrier information, and the path distances is 1. Among them, the weight coefficients of each parameter can be assigned based on the actual situation. For example, the weight coefficient of the computing type is 0.3, the weight coefficient of the current network link status information is 0.2, the weight coefficient of the current computing power resource status information is 0.1, the weight coefficient of the computing node service carrier information is 0.2, and the weight coefficient of the path distance is 0.2. In order to perform weighted calculation on each parameter, each parameter can be evaluated to obtain the score value of each parameter. The score value of each parameter is used to reflect the current status of each parameter, that is, the better the status of each parameter, the lower the score value. The evaluation method can be evaluated by professional technicians based on the actual situation. Among them, the relevant formula for weighted calculation can be shown as follows:

[0062] R = αA + βB + γC + δD + εE;

[0063] In the formula, R represents the weighted calculation result, A represents the score value of the computing type, α represents the weight coefficient of the computing type, B represents the score value of the current network link status information, β represents the weight coefficient of the current network link status information, C represents the score value of the current computing power resource status information, γ represents the weight coefficient of the current computing power resource status information, D represents the score value of the computing node service carrier information, δ represents the weight coefficient of the computing node service carrier information, E represents the score value of the path distance, and ε represents the weight coefficient of the path distance.

[0064] In this embodiment, determining the calculation type, the current network link status information, the current computing power resource status information of all current computing power nodes, the computing power node service carrier information, and the weight coefficient of the path distance includes: if the current network link status information meets the first preset condition, reducing the weight coefficient of the current network link status information to a first preset threshold, and determining the calculation type, the current computing power resource status information of all current computing power nodes, the computing power node service carrier information, and the weight coefficient of the path distance; if the current computing task meets the second preset condition, increasing the weight coefficient of the calculation type to a second preset threshold, and determining the current network link status information, the current computing power resource status information, the computing power node service carrier information, and the weight coefficient of the path distance. Among them, when determining the weight coefficients of various parameters, if the current network link status information indicates that the current network link status is good, that is, the network connection status between each computing power node and its corresponding router is good, and there are fewer packet loss and queuing situations during data transmission, then the first preset condition is met. This means that the parameter of the current network link status information has little impact on data transmission. Therefore, the weight coefficient of the current network link status information can be reduced to 0.1, and the weight coefficients of other parameters can be appropriately increased to select more reasonable computing power nodes; if the current computing task requires a large amount of calculation, it means that its calculation type is calculation-intensive and / or time-sensitive, that is, the second preset condition is met. Therefore, the weight coefficient of the calculation type parameter needs to be increased to the second preset threshold. For example, if the current computing task requires neural network training, that is, a large amount of calculation and higher requirements for the CPU, then the weight coefficient of the calculation type is determined to be the second preset threshold of 0.4. It should be noted that after screening out the current computing power node from all current computing power nodes, it is also necessary to calculate the transmission time of the data packet from the access router to the current computing power node, and then judge whether the current computing power node meets the delay condition of the attribute information in the data packet based on the transmission time of the current computing power node. For example, the delay condition in the attribute information is that the maximum transmission time is 10 ms, and the transmission time from the access router to the current computing power node is 9 ms. Therefore, the delay condition in the attribute information is met, and then the current computing power node is the target computing power node.

[0065] Step S23: If not satisfied, remove the current computing power node from the current set of all computing power nodes to obtain the next set of all computing power nodes, update the next set of all computing power nodes as the current set of all computing power nodes, and then jump back to the step of screening the current computing power node from the current set of all computing power nodes based on the computing type, current network link status information, current computing power resource status information of the current set of all computing power nodes, computing power node service carrier information, and the path distance, until the current computing power node meets the latency condition, so as to determine the current computing power node as the target computing power node.

[0066] In this embodiment, if the current computing power node does not meet the latency condition in the attribute information, for example, the latency condition in the attribute information is that the maximum transmission time is 1 ms, while the transmission time from the access router to the current computing power node is 2 ms, so it does not meet the latency condition in the attribute information. Then the current computing power node cannot be the target computing power node and is removed from the current set of all computing power nodes to obtain the next set of all computing power nodes, and then the current computing power node is screened again until the target computing power node is screened out. It can be understood that computing resources required for the actual computing task can be reserved in the target computing power node, so that the actual computing task can be processed using the resource reservation mechanism later. It should be noted that if multiple computing tasks arrive at the same target computing power node simultaneously, the processing priority can be set for each computing task in advance, so that the target computing power node processes them based on the order of the processing priorities. For example, the processing priority of computing task A is the highest level, the processing priority of computing task B is the middle level, and the processing priority of computing task C is the lowest level. So when computing tasks A, B, and C arrive at target computing power node F simultaneously, the processing order of the target computing power node can be computing tasks A, B, and C. Additionally, if computing task E with the highest processing priority arrives at target computing power node H, and target computing power node H is processing computing task F with the lowest processing priority, then target computing power node H can interrupt the processing of computing task F and process computing task E in the corresponding computing processing unit, so that computing task E does not need to queue and wait.

[0067] Step S24: Based on the target transmission path, use the bandwidth reservation mechanism to transmit the data packet to the target computing power node, so that the target computing power node processes the data packet of the current computing task using the resource reservation mechanism.

[0068] It can be seen that when screening the target computing power node in this application, the weighted calculation method is used, fully considering the influence of various parameters such as the computing type, current network link status information, current computing power resource status information, computing power node service carrier information, and path distance on the transmission and processing of the current computing task, so as to obtain a more reasonable target computing power node, so that the current computing task can be processed with high quality.

[0069] See Figure 3 As shown, an embodiment of the present application discloses a specific method for processing computing tasks, which is applied to a deterministic computing power network system and includes:

[0070] Step S31: Establish a transmission connection between the target terminal and the access router in the deterministic computing power network system, and based on the transmission connection, obtain the data packet of the current computing task of the target terminal and the identification information of the current computing task, and use the identification information to determine the computing type of the current computing task.

[0071] Step S32: Respectively determine the path distances between all computing power nodes and the access router, and based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distances, screen out the target computing power nodes from all computing power nodes.

[0072] Step S33: Plan the transmission path between the target computing power node and the access router to obtain a path library including several such transmission paths, and screen out the current transmission path from the path library in ascending order, and determine whether the task flow of the current transmission path is less than a preset task flow threshold.

[0073] In this embodiment, after screening out the target computing power nodes, a deterministic transmission path is constructed. Specifically, in SDN (Software-defined Networking), the software-defined network controller realizes the control of data plane devices and packet forwarding through the southbound interface protocol; in the deterministic computing power network management plane, the SDN controller function is integrated; on the other hand, SR (Segment Routing) forwards data packets through an ordered instruction list (segment). Therefore, by combining SDN and SR, the SDN controller realizes the planning and decision-making of the routing path, forms a list of SR path information, and then the SDN controller issues a routing instruction with SR path information to the computing network routing device through the southbound interface. The computing network routing device strictly forwards data packets according to the routing instruction, realizing the deterministic control of the transmission path. Using the SDN method and SR technology, first construct all the transmission paths between the target computing power node and the access router, and preferentially screen out the transmission path with the shortest path distance, and judge whether the transmission path reaches the upper limit of the deterministic task flow, that is, it is necessary to judge whether the current bandwidth resource used for transmitting data on the transmission path is less than the preset task flow threshold, so as to ensure that there is sufficient bandwidth resource for transmission when using the transmission path for subsequent transmission. For example, if 3 transmission paths between the target computing power node and the access router are planned, a path library containing transmission path A, transmission path B, and transmission path C is obtained. The transmission paths A, B, and C can be arranged in ascending order. For example, the arranged order is: transmission path A < transmission path C < transmission path B. Screen out the smallest transmission path as the current transmission path, that is, transmission path A, and judge whether the task flow of transmission path A is less than the preset task flow threshold.

[0074] Step S34: If it is less, determine that the current transmission path is the target transmission path; if it is not less, remove the current transmission path from the path library, and then jump back to the step of screening out the current transmission path from several transmission paths in ascending order until the task flow of the current transmission path is less than the preset task flow threshold, so as to obtain the target transmission path.

[0075] It can be understood that in this embodiment, if the task flow of transmission path A is less than the preset task flow threshold, then transmission path A is used as the target transmission path; if transmission path A is not less than the preset task flow threshold, then transmission path A is removed from the path library, and then transmission path C is used as the current transmission path. If the task flow of transmission path C is less than the preset task flow threshold, then transmission path C is the target transmission path, and the bandwidth resource for transmitting the data packet of the current computing task is determined in transmission path C, so as to ensure that there is sufficient bandwidth resource to transmit the data packet and improve the data transmission quality.

[0076] In this embodiment, in order to reduce the time for transmitting the current computing task data packet and the probabilities of packet loss and congestion problems, in addition to adopting the bandwidth resource reservation mechanism, a queue scheduling optimization mechanism can also be combined to ensure the determinacy of transmission delay jitter. The bandwidth resource reservation mechanism means that for a specified transmission path, only the target computing task data packet is set to be transmitted. By enabling the computing task data packet to exclusively occupy the bandwidth resource of the transmission link, congestion problems are ensured not to occur. The queue scheduling optimization mechanism refers to setting a transmission priority for the current computing task. For example, the transmission priority of the current computing task is the highest level. If the data packet of the current computing task conflicts with data packets of other lower levels at a routing and switching node, the data packet of the current computing task can be preferentially transmitted.

[0077] Step S35: Based on the target transmission path, and using the bandwidth reservation mechanism, transmit the data packet to the target computing power node, so that the target computing power node processes the data packet of the current computing task by using the resource reservation mechanism.

[0078] It can be seen that when determining the target transmission path in this application, not only the size of the path needs to be considered, but also whether the task flow of the transmission path is less than a preset task flow threshold needs to be considered, so as to screen out a target transmission path with as short a path distance as possible and a task flow less than the preset task flow threshold. The length of the target transmission path being as short as possible can reduce the data transmission time and improve the transmission efficiency, while the task flow of the target transmission path being less than the preset task flow threshold ensures that there is sufficient broadband resource in the target transmission path to transmit the data packet, achieving high-quality data transmission.

[0079] The following takes Figure 4 a specific schematic diagram of computing task processing shown as an example to illustrate the technical solution in this application. The deterministic computing power network system includes a preset data plane and a preset control plane. Among them, the data plane includes access routers, and the preset control plane includes a computing power service perception unit, a computing power resource perception unit, a network resource perception unit, a network path planning unit, a computing power resource guarantee unit, a network delay control unit, a computing power task scheduling unit, and a network resource guarantee unit.

[0080] The preset data plane establishes a transmission connection between the target terminal and the access router; the access router obtains the data packets of the current computing task and the identification information of the current computing task of the target terminal based on the transmission connection, and reports the data packets and the identification information to the preset management and control plane; the preset management and control plane determines the computing type of the current computing task based on the identification information, determines the path distances between all computing power nodes and the access router, the computing power service perception unit obtains the service carrier information of the computing power nodes, the computing power resource perception unit obtains the current computing power resource status information of all computing power nodes, and the network resource perception unit obtains the current network link status information.

[0081] Based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the service carrier information of the computing power nodes, and the path distances, target computing power nodes are selected from all computing power nodes. During the process of screening the target computing power node network, the network delay control unit determines the transmission time for the data packet to be transmitted from the access router to the current computing power node, and uses the transmission time and the delay condition of the attribute information in the data packet to screen the target computing power nodes. After determining the target computing power nodes, the computing power resource guarantee unit determines the computing resources corresponding to the current computing task reserved in the target computing power nodes.

[0082] The path planning unit uses the software-defined network method and the segment routing technology to plan the target transmission path between the target computing power node and the access router. The network resource guarantee unit determines the broadband resources for transmitting the data packets of the current computing task in the target transmission path to ensure that the data packets have sufficient broadband resources and improve the transmission quality.

[0083] After the data packets are transmitted to the target computing power nodes through the target transmission path, when multiple tasks arrive at the same target computing power node simultaneously, the computing task scheduling unit can determine the processing order of each computing task and can reasonably process the computing tasks.

[0084] See Figure 5 As shown, the embodiment of the present application discloses a computing task processing device, which is applied to a deterministic computing power network system and includes:

[0085] The path distance determination module 11 is used to obtain the data packets of the current computing task, the identification information of the current computing task, and the attribute information of the current computing task, use the identification information to determine the computing type of the current computing task, and then determine the path distances between all computing power nodes and the target terminal;

[0086] The target node screening module 12 is used to select target computing power nodes that meet the preset computing power node computing resource reservation mechanism from all computing power nodes based on the attribute information, the current network link status information, the current computing power resource status information of all computing power nodes, and the path distances;

[0087] The processing module 13 is configured to plan a target transmission path between the target computing power node and the target terminal, and transmit the data packet to the target computing power node based on the target transmission path, so that the target computing power node processes the current computing task based on the data packet.

[0088] Among them, the more specific working processes of the above-mentioned various modules can refer to the corresponding content disclosed in the foregoing embodiments, and will not be elaborated herein.

[0089] It can be seen that this application establishes a transmission connection between the target terminal and the access router in the deterministic computing power network system, and obtains the data packet of the current computing task and the identification information of the current computing task of the target terminal based on the transmission connection, and determines the computing type of the current computing task by using the identification information; respectively determines the path distances between all computing power nodes and the access router, and based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distances, filters out the target computing power node from all computing power nodes, and uses software-defined network methods and segment routing technologies to plan the target transmission path between the target computing power node and the access router; based on the target transmission path, and uses the bandwidth reservation mechanism to transmit the data packet to the target computing power node, so that the target computing power node uses the resource reservation mechanism to process the data packet of the current computing task. Thus, it can be seen that when determining the target computing power node in this application, it is necessary to take into account the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distances, so that the selected target computing power node can meet the network link status, computing power resource status, service carrier, computing type, and path distances required by the current computing task, and performs resource reservation processing on the target computing power node, that is, the subsequent target computing power node uses the resource reservation mechanism to process the data packet of the current computing task, thereby ensuring that the target computing power node has sufficient computing resources to process the current computing task, that is, ensuring that the current computing task can be processed with high quality; when planning the target transmission path between the target computing power node and the access router, software-defined network methods and segment routing technologies are used, so that the target transmission path is the optimal among all transmission paths, and because the broadband reservation mechanism is used when transmitting the data packet of the current computing task, it is ensured that the target transmission path has sufficient broadband resources to transmit the data packet of the current computing task to the target computing power node, that is, the data packet of the current computing task can be transmitted to the target computing power node with high quality.

[0090] Furthermore, an embodiment of this application also provides an electronic device. Figure 6It is a structural diagram of an electronic device 20 shown according to an exemplary embodiment. The content in the figure should not be considered as any limitation on the scope of use of this application.

[0091] Figure 6 It is a schematic structural diagram of an electronic device provided by an embodiment of this application. Specifically, it may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. Among them, the memory 22 is used to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the relevant steps in the computing task processing method executed by the electronic device disclosed in any of the foregoing embodiments.

[0092] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device; the communication interface 24 can create a data transmission channel between the electronic device and external devices, and the communication protocol it follows is any communication protocol applicable to the technical solution of this application, and no specific limitation is imposed on it here; the input / output interface 25 is used to obtain external input data or output data to the outside, and the specific interface type can be selected according to specific application needs, and no specific limitation is made here.

[0093] Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, and the AI processor is used to process computing operations related to machine learning.

[0094] In addition, as a carrier for storing resources, the memory 22 can be a read-only memory, a random access memory, a magnetic disk, an optical disc, etc. The resources stored thereon include an operating system 221, a computer program 222, data 223, etc. The storage method can be transient storage or permanent storage.

[0095] Among them, the operating system 221 is used to manage and control each hardware device and the computer program 222 on the electronic device, so as to implement the operation and processing of the massive data 223 in the memory 22 by the processor 21. It can be Windows, Unix, Linux, etc. In addition to the computer program that can be used to complete the calculation task processing method executed by the electronic device disclosed in any of the foregoing embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks. The data 223 can include not only the data transmitted by the external device received by the electronic device, but also the data collected by its own input / output interface 25, etc.

[0096] Furthermore, an embodiment of the present application also discloses a computer-readable storage medium. When the computer program stored in the storage medium is loaded and executed by a processor, the method steps executed during the calculation task processing disclosed in any of the foregoing embodiments are implemented.

[0097] Finally, it should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0098] The above has introduced in detail a calculation task processing method, device, equipment and medium provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for processing computing tasks, characterized in that, Applied to a deterministic computing power network system, including: Establish a transmission connection between the target terminal and the access router in the deterministic computing power network system, and based on the transmission connection, obtain the data packet of the current computing task of the target terminal and the identification information of the current computing task, and use the identification information to determine the computing type of the current computing task; Respectively determine the path distances between all computing power nodes and the access router, and based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distances, screen out the target computing power nodes from all computing power nodes, and use software-defined network methods and segment routing technologies to plan the target transmission path between the target computing power nodes and the access router; Based on the target transmission path, and use the bandwidth reservation mechanism to transmit the data packet to the target computing power node, so that the target computing power node uses the resource reservation mechanism to process the data packet of the current computing task; Among them, the screening out of the target computing power nodes from all computing power nodes based on the computing type, the current network link status information, the current computing power resource status information of all computing power nodes, the computing power node service carrier information, and the path distances includes: Determine the weight coefficients and scoring values of the computing type, the current network link status information, the current computing power resource status information of all current computing power nodes, the computing power node service carrier information, and the path distances; use the weight coefficients to perform weighted calculations on the scoring values to obtain the weighted calculation results of all current computing power nodes, and screen out the current computing power nodes based on the order from large to small; calculate the transmission time of the data packet from the access router to the current computing power node, and judge whether the current computing power node meets the delay condition of the attribute information in the data packet based on the transmission time of the current computing power node; if the delay condition is met, determine the current computing power node as the target computing power node; The determination of the weight coefficients of the computing type, the current network link status information, the current computing power resource status information of all current computing power nodes, the computing power node service carrier information, and the path distances includes: If the current network link status information meets the first preset condition, reduce the weight coefficient of the current network link status information to the first preset threshold, and determine the weight coefficients of the computing type, the current computing power resource status information of all current computing power nodes, the computing power node service carrier information, and the path distances; if the current computing task meets the second preset condition, increase the weight coefficient of the computing type to the second preset threshold, and determine the weight coefficients of the current network link status information, the current computing power resource status information, the computing power node service carrier information, and the path distances.

2. The calculation task processing method according to claim 1, wherein Before obtaining the data packet of the current computing task of the target terminal and the identification information of the current computing task based on the transmission connection, it further includes: If the type of the current computing task is compute-intensive, set the preset identification position of the current computing task through the target terminal to obtain the identification information of the current computing task corresponding to the compute-intensive type; and / or, if the type of the current computing task is time-sensitive, set the preset identification position of the current computing task through the target terminal to obtain the identification information of the current computing task corresponding to the time-sensitive type.

3. The computing task processing method according to claim 2, wherein, Before screening out the target computing nodes from all the computing nodes based on the computing type, the current network link state information, the current computing resource state information of all the computing nodes, the computing node service carrier information, and the path distance, it further includes: Obtain the current network link state information, the computing node service carrier information, and the current computing resource state information including the current load information of all the computing nodes.

4. The method for processing a computing task according to any one of claims 1 to 3, characterized in that, The screening out of the target computing nodes from all the computing nodes based on the computing type, the current network link state information, the current computing resource state information of all the computing nodes, the computing node service carrier information, and the path distance further includes: If not satisfied, remove the current computing node from the current all computing nodes to obtain the next all computing nodes, update the next all computing nodes to the current all computing nodes, and re-jump to the step of screening out the current computing node from the current all computing nodes based on the computing type, the current network link state information, the current computing resource state information of the current all computing nodes, the computing node service carrier information, and the path distance until the current computing node meets the latency condition, so as to determine the current computing node as the target computing node.

5. The calculation task processing method according to claim 1, wherein The planning of the target transmission path between the target computing node and the access router by using the software-defined network method and the segment routing technology includes: Plan the transmission path between the target computing node and the access router to obtain a path library containing several such transmission paths, and screen out the current transmission path from the path library in ascending order, and determine whether the task flow of the current transmission path is less than the preset task flow threshold; If it is less, determine the current transmission path as the target transmission path. If it is not less, remove the current transmission path from the path library, and re-jump to the step of screening out the current transmission path from several such transmission paths in ascending order until the task flow of the current transmission path is less than the preset task flow threshold, thereby obtaining the target transmission path.

6. A computing task processing device, characterized in that, Applied to a deterministic computing power network system, it includes: A path distance determination module, configured to obtain the data packet of the current computing task, the identification information of the current computing task, and the attribute information of the current computing task, and use the identification information to determine the computing type of the current computing task, and determine the path distance between all the computing nodes and the target terminal; A target node screening module, configured to screen out target computing nodes that meet the preset computing resource reservation mechanism of computing nodes from all the computing nodes based on the attribute information, the current network link status information, the current computing resource status information of all the computing nodes, the computing node service carrier information, and the path distance; A processing module, configured to plan a target transmission path between the target computing node and the target terminal, and transmit the data packet to the target computing node based on the target transmission path, so that the target computing node processes the data packet of the current computing task by using the resource reservation mechanism; Wherein, the target node screening module specifically includes: Determine the weight coefficients and scoring values of the computing type, the current network link status information, the current computing resource status information of all current computing nodes, the computing node service carrier information, and the path distance; perform weighted calculation on the scoring values by using the weight coefficients to obtain the weighted calculation results of all current computing nodes, and screen out the current computing nodes from the weighted calculation results in descending order; calculate the transmission time of the data packet from the access router to the current computing node, and determine whether the current computing node meets the delay condition of the attribute information in the data packet based on the transmission time of the current computing node; if the delay condition is met, determine the current computing node as the target computing node; The target node screening module specifically includes: If the current network link status information meets the first preset condition, reduce the weight coefficient of the current network link status information to a first preset threshold, and determine the weight coefficients of the computing type, the current computing resource status information of all current computing nodes, the computing node service carrier information, and the path distance; if the current computing task meets the second preset condition, increase the weight coefficient of the computing type to a second preset threshold, and determine the weight coefficients of the current network link status information, the current computing resource status information, the computing node service carrier information, and the path distance.

7. An electronic device, characterized in that, It includes: A memory, configured to store a computer program; A processor, configured to execute the computer program to implement the steps of the computing task processing method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, For storing a computer program; wherein, when the computer program is executed by the processor, the steps of the computing task processing method according to any one of claims 1 to 5 are implemented.

Images