Intelligent computing virtualization computing power distribution method, structure and device based on V2V

Abstract

The embodiment of the invention provides a V2V-based intelligent computing virtualization computing power distribution method, structure and device, a VRB network architecture comprises a plurality of computing power devices and a VRB switch, and the computing power devices are connected through the VRB switch; the method comprises the following steps: acquiring equipment information of a plurality of computing power equipment in a VRB network architecture; obtaining a calculation task, and determining a target computing power device according to the calculation task and the device information of the computing power device; the target computing power device is scheduled to execute the computing task, and data involved in the process of executing the computing task by the target computing power device is transmitted through the VRB switch based on the VRB protocol; according to the computing power scheduling method based on the VRB network architecture, it is guaranteed that the scheduled computing power task can make full use of the powerful underlying communication capacity provided by the V2V protocol and the VRB architecture, and the data transmission efficiency and reliability between computing power resources are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and particularly to a V2V-based intelligent computing virtualized computing power allocation method, structure, and device. Background Art

[0002] With the continuous increase in the complexity of intelligent computing tasks and the scale of data sets, it has become inevitable to allocate and schedule computing tasks to different computing power resources for collaborative operation; however, when the current network architecture processes complex tasks such as large-scale deep learning training and real-time analysis of massive data, the data transmission delay is large and the stability is poor under the traditional architecture, which cannot ensure the timely and accurate interaction of key data, severely restricting the improvement of the overall computing efficiency. Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention are proposed to provide a V2V-based intelligent computing virtualized computing power allocation method, structure, device, and equipment that overcome the above problems or at least partially solve the above problems.

[0004] To solve the above problems, embodiments of the present invention disclose a V2V-based intelligent computing virtualized computing power allocation method, which is applied to a VRB network architecture. The VRB network architecture includes multiple computing power devices and a VRB switch, and the computing power devices are connected through the VRB switch; the method includes:

[0005] Obtain the device information of multiple computing power devices in the VRB network architecture;

[0006] Obtain a computing task, and determine a target computing power device according to the computing task and the device information of the computing power device;

[0007] Schedule the target computing power device to execute the computing task, and transmit the data involved in the process of the target computing power device executing the computing task through the VRB switch based on the VRB protocol.

[0008] Optionally, the multiple computing power devices are deployed in multiple clusters, and the computing power devices in the same cluster are connected through the same VRB switch; the computing power devices between different clusters are connected through the VRB switches corresponding to different clusters;

[0009] The transmitting the data involved in the process of the target computing power device executing the computing task through the VRB switch based on the VRB protocol includes:

[0010] Transmit the data involved in the process of the target computing power device in the same cluster executing the computing task through the VRB switch corresponding to the cluster based on the VRB protocol;

[0011] Through the VRB switches corresponding to different clusters, transmit the data involved in the process of the target computing devices in different clusters executing the computing tasks based on the VRB protocol.

[0012] Optionally, multiple computing devices in the VRB network architecture are respectively connected to the V2V switch. The determining of the target computing device according to the computing task and the device information of the computing device includes:

[0013] Determine the initial computing device according to the computing task and the device information of the computing device;

[0014] Obtain the operating status of the initial computing device through the V2V switch;

[0015] Determine the candidate computing devices according to the operating status of the initial computing device;

[0016] Determine the target computing device among the candidate computing devices.

[0017] Optionally, the obtaining of the device information of the computing devices in the VRB network architecture includes:

[0018] Obtain the device information of each computing device in the VRB network architecture through the V2V switch.

[0019] Optionally, the obtaining of the operating status of the initial computing device through the V2V switch includes:

[0020] Send a query request to the initial computing device through the V2V switch, and the query request is used to obtain the operating status;

[0021] Receive the operating status sent by the initial computing device through the V2V switch.

[0022] Optionally, the computing task includes multiple subtasks and the priorities of each subtask among the multiple subtasks;

[0023] The scheduling of the target computing device to execute the computing task includes:

[0024] Determine the order of the target computing devices corresponding to each subtask to execute the tasks according to the priorities of each subtask;

[0025] Send the multiple subtasks to the target computing devices corresponding to each subtask in sequence through the V2V switch.

[0026] Optionally, the determining of the target computing device among the candidate computing devices includes:

[0027] Determine the available resource amounts of each computing power device among the candidate computing power devices;

[0028] Sort each of the computing power devices according to the available resource amounts of the computing power devices to obtain a sorting result;

[0029] Determine the computing power device with the largest available resource amount in the sorting result as the target computing power device.

[0030] Optionally, the determining the initial computing power device according to the computing task and the device information of the computing power device includes:

[0031] Construct a computing power resource pool according to the device information of the computing power device, where the computing power resource pool includes the device information corresponding to different computing power devices;

[0032] Query the initial computing power device corresponding to the computing task in the computing power resource pool.

[0033] Optionally, the determining the candidate computing power devices according to the operating status of the initial computing power device includes:

[0034] Determine the computing power devices with an idle operating status among the initial computing power devices as the candidate computing power devices.

[0035] The present invention also discloses an intelligent computing virtualized computing power allocation structure based on V2V. The structure includes a task scheduling platform and a VRB network architecture. The VRB network architecture includes multiple computing power devices and a VRB switch, and the computing power devices are connected through the VRB switch;

[0036] The task scheduling platform is configured to obtain the device information of multiple computing power devices in the VRB network architecture; obtain a computing task, determine a target computing power device according to the computing task and the device information of the computing power device; schedule the target computing power device to execute the computing task;

[0037] The VRB switch is configured to transmit the data involved in the process of the target computing power device executing the computing task based on the VRB protocol.

[0038] Optionally, the multiple computing power devices are deployed in multiple clusters, and the computing power devices in the same cluster are connected through the same VRB switch; the computing power devices between different clusters are connected through the VRB switches corresponding to different clusters;

[0039] The VRB switch corresponding to the same cluster is configured to transmit the data involved in the process of the target computing power device in the same cluster executing the computing task based on the VRB protocol;

[0040] The VRB switches corresponding to different clusters are used to transmit data involved in the process of the target computing devices in different clusters executing the computing tasks based on the VRB protocol.

[0041] Optionally, multiple computing devices in the VRB network architecture are respectively connected to the V2V switch, and the task scheduling platform is used to determine the initial computing devices according to the computing tasks and the device information of the computing devices.

[0042] The V2V switch is used to obtain the operating status of the initial computing devices and send it to the task scheduling platform.

[0043] The task scheduling platform is used to determine candidate computing devices according to the operating status of the initial computing devices; and determine the target computing devices among the candidate computing devices.

[0044] Optionally, the V2V switch is used to obtain the device information of each computing device in the VRB network architecture and send it to the task scheduling platform.

[0045] Optionally, the computing task includes multiple subtasks and the priorities of each subtask among the multiple subtasks.

[0046] The task scheduling platform is used to determine the order of the target computing devices corresponding to each subtask to execute the tasks according to the priorities of each subtask.

[0047] The V2V switch is used to send the multiple subtasks to the target computing devices corresponding to each subtask in accordance with the order.

[0048] Optionally, the task scheduling platform is used to determine the available resource amounts of each computing device among the candidate computing devices; sort each computing device according to the available resource amounts of each computing device to obtain a sorting result; and determine the computing device with the largest available resource amount in the sorting result as the target computing device.

[0049] Optionally, the task scheduling platform is used to construct a computing resource pool according to the device information of the computing devices, where the computing resource pool includes the device information corresponding to different computing devices; and query the initial computing devices corresponding to the computing tasks in the computing resource pool.

[0050] Optionally, the task scheduling platform is used to determine the computing devices with an idle operating status among the initial computing devices as candidate computing devices.

[0051] The present invention also discloses an intelligent computing virtualized computing power allocation device based on V2V, which is applied to a VRB network architecture. The VRB network architecture includes multiple computing power devices and a VRB switch, and the computing power devices are connected through the VRB switch; the device includes:

[0052] A first acquisition module, configured to acquire device information of multiple computing power devices in the VRB network architecture;

[0053] A second acquisition module, configured to acquire a computing task, and determine a target computing power device according to the computing task and the device information of the computing power device;

[0054] A scheduling module, configured to schedule the target computing power device to execute the computing task, and transmit data involved in the process of the target computing power device executing the computing task through the VRB switch based on the VRB protocol.

[0055] The present invention also discloses an electronic device, including: a processor, a memory, and a computer program stored on the memory and capable of running on the processor. When the computer program is executed by the processor, the steps of the above-mentioned intelligent computing virtualized computing power allocation method based on V2V are implemented.

[0056] The present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned intelligent computing virtualized computing power allocation method based on V2V are implemented.

[0057] The embodiments of the present invention have the following advantages:

[0058] The present invention discloses an intelligent computing virtualized computing power allocation method based on V2V. The present invention can use a VRB switch to transmit data involved in the process of a target computing power device executing a computing task based on the VRB protocol. The VRB protocol is usually designed to meet the requirements of efficient data transmission under a specific network architecture, and can ensure that data is quickly and accurately transmitted between computing power devices, reducing data transmission delay and packet loss rate, and greatly improving the data transmission efficiency and reliability between computing power resources; since the VRB protocol communicates based on the visual networking protocol, the communication node exchange mechanism relying on the V2V protocol can optimize the data flow path, ensuring efficient cooperation of computing power resources and stable data interaction from the communication basic level, and reducing the cumbersome processes brought by intermediate nodes in the traditional multi-layer addressing method; the present invention can effectively solve the problems existing in the data transmission process of computing power resources in the traditional network architecture, removing obstacles for data transmission between computing power resources; it can dynamically determine the target computing power device according to actual task requirements, realizing rapid and accurate positioning of the target computing power device, and improving the efficiency of establishing a communication link. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of the steps of a V2V-based intelligent computing virtualized computing power allocation method provided by an embodiment of the present invention;

[0060] Figure 2 It is a schematic diagram of a computing power allocation scenario provided by an embodiment of the present invention;

[0061] Figure 3 It is a flowchart of a V2V-based intelligent computing virtualized computing power allocation method provided by an embodiment of the present invention;

[0062] Figure 4 It is a block diagram of the structure of a V2V-based intelligent computing virtualized computing power allocation structure provided by an embodiment of the present invention;

[0063] Figure 5 It is a block diagram of the structure of a V2V-based intelligent computing virtualized computing power allocation device provided by an embodiment of the present invention. Detailed implementation manners

[0064] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0065] Traditional network architectures often adopt a hierarchical topology structure, such as a core-aggregation-access three-layer architecture. In this structure, when data flows from the access layer to the core layer, it needs to be forwarded through multiple intermediate nodes. For the large amount of data interaction in high-performance collaborative computing, this multi-hop forwarding will introduce a large delay and affect real-time performance. Moreover, traditional networks mainly communicate based on the TCP / IP protocol. Although the TCP protocol has a reliability guarantee mechanism, in high-performance collaborative computing scenarios, its slow start, congestion control, and other mechanisms will limit the speed and real-time performance of data transmission. For example, when performing large-scale data parallel computing, a large amount of intermediate results need to be quickly transmitted between multiple computing devices, and the congestion control of TCP will cause the data transmission speed to fail to meet real-time requirements.

[0066] One of the core concepts of the embodiments of the present invention is that the present invention can use a VRB switch to transmit data involved in the process of a target computing device executing a computing task based on the VRB protocol. The VRB protocol is usually designed to meet the requirements of efficient data transmission under a specific network architecture, which can ensure the rapid and accurate transmission of data between computing devices, reduce data transmission latency and packet loss rate, and greatly improve the data transmission efficiency and reliability between computing resources. Since the VRB protocol communicates based on the Visual Networking Protocol, the communication node exchange mechanism relying on the V2V protocol can optimize the data flow path, ensuring efficient collaboration of computing resources and stable data interaction at the communication infrastructure level, and reducing the cumbersome processes brought by intermediate nodes in the traditional multi-layer addressing method. The present invention can effectively solve the problems existing in the data transmission process of computing resources in the traditional network architecture, removing obstacles for data transmission between computing resources. It can dynamically determine the target computing device according to the actual task requirements, quickly and accurately locate the target computing device, and improve the efficiency of establishing a communication link.

[0067] Referring to Figure 1 , a structural block diagram of an intelligent computing virtualized computing power allocation method based on V2V provided by an embodiment of the present invention is shown. The VRB network architecture includes multiple computing devices and a VRB switch, and the computing devices are connected through the VRB switch. The method may include the following steps:

[0068] Step 101, obtain the device information of multiple computing devices in the VRB network architecture.

[0069] In the embodiments of the present invention, VRB is an abbreviation for V2V RDMA Band, which is a network protocol that allows remote direct memory access through a V2V network. It enables low-latency and high-bandwidth data transmission in a V2V network environment, and can effectively improve the data communication performance in an intelligent computing data center environment.

[0070] The computing device may be at least one of a CPU server and a GPU server. The device information may include computing power attributes information such as computing ability (such as the main frequency and number of cores of the CPU, the floating-point computing ability and video memory capacity of the GPU), memory capacity, network bandwidth, and device identification information.

[0071] These information can be collected through a dedicated management software or monitoring system. Information request instructions can be sent to each computing device regularly or in real time. After receiving the request, the computing device will feedback its own device information to the management system. Or when the computing device starts up or connects to the network architecture, it can actively report its own information to the management system. For example, in a data center, each computing device can send a data packet containing its own hardware information and initial status information to the central management system when starting up, so as to ensure that the management system is aware of its performance and availability.

[0072] Step 102: Obtain a computing task, and determine a target computing device according to the computing task and the device information of the computing power device.

[0073] In the embodiment of the present invention, the computing task is used to represent the relevant description information of the task, which clarifies at least one of the following information: the type of the task (such as data processing task, image processing task, scientific computing task or other types), the required amounts of various computing resources, the location requirements for data storage, the bandwidth and latency requirements for network communication, the approximate range of the expected execution time, and the priority of the task.

[0074] When determining the target computing device, the required resources of the task can be matched with the available resources of the computing power device. For example, for a deep learning training task that requires a large amount of GPU resources, computing power devices with high-performance GPUs and low current GPU loads will be given priority; for tasks with high requirements for storage speed, they will be assigned to computing power devices with fast storage systems. At the same time, the priority of the task will also be considered. High-priority tasks may be assigned to the devices with the best performance and the lightest load to ensure their completion as soon as possible. This process can be implemented using some algorithms, such as the greedy algorithm, genetic algorithm or rule-based algorithm, to find the optimal target computing device according to different resource allocation strategies.

[0075] Step 103: Schedule the target computing device to execute the computing task, and transmit the data involved in the process of the target computing device executing the computing task through a VRB switch based on the VRB protocol.

[0076] In the embodiment of the present invention, after determining the target computing device, the computing task can be sent to the target computing device, and a corresponding start instruction can be sent to the computing power device to make it start processing the task. In terms of data transmission, when the target computing device starts to execute the task, it will involve the reading of input data, the transmission and storage of intermediate results, and the output of the final result. The VRB switch is responsible for completing the transmission of these data in the entire network architecture according to the VRB protocol. For example, when the target computing device needs to obtain the data of the previous computing power device, the data can be encapsulated according to the VRB protocol and transmitted to the target computing device through the VRB switch; during the computing process, the intermediate results at different stages may be transmitted between different computing power devices, and the VRB switch routes and forwards these data according to the protocol.

[0077] The present invention discloses a V2V-based intelligent computing virtualized computing power allocation method. The present invention can utilize a VRB switch to transmit data involved in the process of a target computing power device executing a computing task based on the VRB protocol. The VRB protocol is usually designed to meet the requirements of efficient data transmission under a specific network architecture, which can ensure the rapid and accurate transmission of data between computing power devices, reduce data transmission latency and packet loss rate, and greatly improve the data transmission efficiency and reliability between computing power resources. Since the VRB protocol communicates based on the Visual Networking Protocol, the communication node exchange mechanism relying on the V2V protocol can optimize the data flow path, ensuring efficient cooperation of computing power resources and stable data interaction at the communication foundation level, and reducing the cumbersome processes brought by intermediate nodes in the traditional multi-layer addressing method. The present invention can effectively solve the problems existing in the data transmission process of computing power resources in the traditional network architecture, removing obstacles for data transmission between computing power resources, and can dynamically determine the target computing power device according to actual task requirements, achieving rapid and accurate positioning of the target computing power device and improving the efficiency of establishing a communication link.

[0078] In an embodiment of the present invention, multiple computing power devices are deployed in multiple clusters. The computing power devices in the same cluster are connected through the same VRB switch, and the computing power devices between different clusters are connected through the VRB switches corresponding to different clusters.

[0079] Through the VRB switch, transmit the data involved in the process of the target computing power device executing a computing task based on the VRB protocol, including: transmitting the data involved in the process of the target computing power device in the same cluster executing a computing task based on the VRB protocol through the VRB switch corresponding to the cluster; transmitting the data involved in the process of the target computing power device in different clusters executing a computing task based on the VRB protocol through the VRB switches corresponding to different clusters.

[0080] In an embodiment of the present invention, as Figure 2 , it shows a schematic diagram of a computing power allocation scenario provided by an embodiment of the present invention. Multiple computing power devices are deployed in multiple clusters. Among them, the CPU server cluster includes multiple CPU servers, the GPU server cluster includes multiple GPU servers. The CPU servers are connected through a VRB switch, the GPU servers are connected through a VRB switch, and the CPU server cluster and the GPU server cluster are communicatively connected through VRB switch 1 and VRB switch 2.

[0081] When the computing power devices in the same cluster need to transmit data, for example, when the CPU servers in the CPU server cluster need to transmit data, the data involved in the computing task can be transmitted through VRB switch 1. When the GPU servers in the GPU server cluster need to transmit data, the data involved in the computing task can be transmitted through VRB switch 2.

[0082] When a CPU server in a CPU cluster needs to transfer data to a GPU server in a GPU server cluster, the CPU server can first transfer the involved data to VRB Switch 1, then VRB transfers the involved data to VRB Switch 2, and then VRB Switch 2 transfers the involved data to the GPU server.

[0083] In an embodiment of the present invention, multiple computing power devices in the VRB network architecture are respectively connected to the V2V switch. According to the computing task and the device information of the computing power device, determining the target computing power device includes: determining the initial computing power device according to the computing task and the device information of the computing power device; obtaining the running state of the initial computing power device through the V2V switch; determining the candidate computing power device according to the running state of the initial computing power device; and determining the target computing power device among the candidate computing power devices.

[0084] In the embodiment of the present invention, as Figure 2 , multiple computing power devices in the VRB network architecture are respectively connected to the V2V switch. In one example, such as the type of computing task (such as scientific computing, data processing, graphics rendering, etc.), the required computing resources (such as the number of CPU cores, memory capacity, storage requirements, GPU performance requirements, etc.). Then, these task information are matched with the device information of each computing power device. The device information may include the hardware configuration of the computing power device (such as processor type and performance, memory size, storage capacity, GPU specifications, etc.), the current load condition, the historical performance data of the device, etc. For each computing power device, it is judged whether it meets the basic requirements of the computing task, such as whether the number of processor cores of the computing power device meets the requirements of the task, whether the memory is sufficient, whether the storage can accommodate the required data, etc. The computing power devices that meet the basic requirements will be listed as the initial computing power devices. For example, if the computing task requires a large amount of parallel computing power, then those computing power devices with multi-core high-performance processors and relatively low current loads may be preferentially considered as the initial computing power devices.

[0085] As a key component connecting multiple computing power devices, the V2V switch has the ability to monitor and manage the connected devices. It can send status query requests to the initial computing power devices. These requests may include the current CPU usage rate, memory occupancy rate, network bandwidth usage, other task information being processed, device temperature, hardware health status, etc. of the device. After receiving the requests, the initial computing power devices will feedback their real-time running state information to the V2V switch, and then the V2V switch forwards it to the corresponding management or scheduling system. For example, an initial computing power device may seem to have sufficient computing resources in the previous screening, but after obtaining its running state through the V2V switch, it is found that its CPU usage rate is close to saturation, so it may not be suitable for immediately executing a new computing task.

[0086] For the operating status information of each initial computing power device obtained from the V2V switch, a detailed evaluation will be carried out. For example, if the CPU utilization rate of an initial computing power device is too high (such as exceeding 80%), or the device temperature is too high, it may be excluded from the candidate computing power devices because they may not be able to provide sufficient resources for the current computing task or may have a risk of failure during task execution. While those initial computing power devices with good operating status and sufficient resource margins (such as CPU utilization rate below 50%, sufficient available space in memory and storage, etc.) will be determined as candidate computing power devices. At the same time, for some computing tasks with high requirements for task execution time, the historical performance data of the devices will also be considered, and those devices that have shown higher efficiency and stability when performing similar tasks in the past will be preferentially selected as candidate computing power devices.

[0087] The target computing power device can be selected from the candidate computing power devices according to different strategies. A common strategy is based on performance optimization, that is, to select the candidate computing power device with the highest available resource margin and the best performance matching the task requirements. For example, for a computing task that requires high GPU performance, the candidate computing power device with the strongest GPU and the largest GPU resource margin will be selected. Another strategy can be based on cost or energy consumption considerations. On the premise of meeting the task requirements, select those candidate computing power devices that consume fewer resources or have lower energy consumption. In addition, according to the load balancing principle, the candidate computing power device with the lightest load can be preferentially selected to avoid a certain device being in a high-load state for a long time and improve the stability and resource utilization rate of the entire VRB network architecture. If there are multiple candidate computing power devices that meet the conditions, some algorithms, such as random algorithms or weighted algorithms, may be used to evaluate the candidate computing power devices according to different weights, and finally determine the target computing power device. Which strategy to adopt specifically can be set according to user needs.

[0088] In an embodiment of the present invention, the data involved in the computing task process can be added with verification information through the VRB switch, and the data added with verification information is transmitted based on the VRB protocol.

[0089] In the embodiments of the present invention, when there is data that needs to be transmitted between computing power servers, each computing power server can utilize the efficient communication mechanism of the VRB network architecture to quickly establish a direct data transmission link with the target server. It can send the data to be transmitted to the target computing power server through the V2V network according to a predetermined protocol, adding necessary verification information and transmission control information through the VRB network communication protocol and the data transmission interface. After receiving the data, the target computing power server reorganizes and verifies the data according to the transmission control information to ensure the integrity and accuracy of the data, and directly provides the received data to the task process being executed for further processing. In this process, the data flows quickly and accurately between the computing power servers according to the task requirements and the communication capabilities of the VRB network, ensuring the smooth progress of the collaborative work during the task execution, avoiding task delays or failures caused by data transmission problems. The flow direction and transmission status of the data are monitored and managed by each server and the VRB network throughout the process to ensure the efficiency and reliability of data interaction.

[0090] In an embodiment of the present invention, obtaining the device information of computing power devices in the VRB network architecture includes: obtaining the device information of each computing power device in the VRB network architecture through a V2V switch.

[0091] In the embodiments of the present invention, as Figure 2 , the computing power scheduling platform can obtain the device information of each computing power device in the VRB network architecture through a V2V switch, and can converge the connections between different computing power devices to coordinate the flow of data and the interaction between devices.

[0092] In an embodiment of the present invention, obtaining the operating status of the initial computing power device through a V2V switch includes: sending a query request to the initial computing power device through the V2V switch, where the query request is used to obtain the operating status; receiving the operating status sent by the initial computing power device through the V2V switch.

[0093] In the embodiments of the present invention, the V2V switch can send a device information request message to each computing power device. This request message can be a specific protocol packet containing an instruction to request the acquisition of device information. When the computing power device receives this message, it will organize the required device information into a corresponding message format according to its own hardware and software status.

[0094] For example, the CPU computing power device will package its own CPU information, memory information, storage information, etc. into a data frame, which includes the identifier of the device, the specific data of each item of information, and the corresponding information type (such as the identification information is CPU information, storage information, etc.).

[0095] The computing device sends the organized device information message back to the V2V switch. The V2V switch receives information from each computing device, classifies, stores, and manages this information according to the device identifier for subsequent use.

[0096] In an embodiment of the present invention, the computing task includes multiple subtasks and the priority of each subtask among the multiple subtasks; scheduling the target computing device to execute the computing task includes: determining the order in which the target computing devices corresponding to each subtask execute the task according to the priority of each subtask; and sending the multiple subtasks to the target computing devices corresponding to each subtask in sequence through the V2V switch.

[0097] In the embodiment of the present invention, to ensure that the subtasks in the computing task can be executed on the target computing device in an orderly manner according to importance or urgency to meet business requirements and improve the overall computing efficiency, by reasonably arranging the execution order, key subtasks can be processed first, avoiding delays in high-priority tasks caused by low-priority tasks occupying resources.

[0098] When the computing task includes multiple subtasks, when the computing task is configured, a priority identifier will be assigned to each subtask. This priority identifier can be a numerical value (for example, 1 represents the highest priority, and the larger the value, the lower the priority), or a specific priority label (such as "high", "medium", "low"). The determination of the priority is usually based on business logic and the characteristics of the computing task. For example, in a real-time data analysis scenario, subtasks related to real-time decision-making may be assigned a high priority, while some auxiliary data cleaning subtasks have a relatively low priority.

[0099] A sorting algorithm can be used to sort the subtasks according to their priorities. Taking bubble sort as an example, it will compare the priorities of adjacent subtasks in sequence. If the order is incorrect, they will be swapped. After multiple comparisons and swaps, finally all subtasks will be arranged in descending order of priority.

[0100] Before sending the subtasks, each subtask can be encapsulated to form a data packet in a specific format. This data packet, in addition to containing the specific computing content of the subtask, will also add metadata such as the identification information of the target computing device and the sequence number of the subtask in the entire task sequence. For example, the data packet has a header and a data part. The header contains information such as the IP address of the target computing device and the subtask sequence number, and the data part is the actual computing code and related data of the subtask.

[0101] V2V switch forwarding: The encapsulated subtask data packet is sent to the V2V switch. The V2V switch can search its internal routing table or forwarding rules based on the target computing device identification information in the packet header. The routing table or forwarding rules record the connection information and forwarding path of each computing device. For example, when the V2V switch receives a subtask data packet whose target is computing device X, it will find the port information leading to computing device X from the routing table, and then forward the data packet from the port to ensure that the data packet can accurately reach the corresponding target computing device.

[0102] In order to ensure that the subtasks are sent in the predetermined order, the V2V switch can adopt a queue mechanism. It can create a sending queue for each subtask, and put the encapsulated data packets into the queue in sequence according to the execution order determined by the subtasks. The V2V switch takes out the data packets from the head of the queue in sequence for forwarding, which ensures that the subtasks can be sent to the target computing device in sequence. At the same time, in order to prevent the data packet of a high-priority subtask from occupying the queue for a long time, causing the low-priority subtask to starve, some scheduling algorithms can be used to reasonably allocate sending resources, such as time slice round-robin scheduling to reasonably allocate sending resources to ensure that each subtask has the opportunity to be sent.

[0103] In one embodiment of the present invention, on the basis of simply sorting by priority, the order can also be fine-tuned in combination with the resource status and current load status of the target computing device. For example, if a high-priority subtask has a huge demand for GPU resources, and the GPU of the current target computing device A is busy processing other tasks and has tight resources, while the GPU resources of the target computing device B are relatively abundant, even if the subtask originally corresponds to device A in the priority sorting, it may be adjusted to let device B execute the subtask first to ensure that the high-priority subtask can be processed as soon as possible.

[0104] In one embodiment of the present invention, a target computing power device is determined among candidate computing power devices, including: determining the amount of available resources of each computing power device among the candidate computing power devices; sorting each computing power device according to the amount of available resources of each computing power device to obtain a sorting result; and determining the computing power device with the largest amount of available resources in the sorting result as the target computing power device.

[0105] In an embodiment of the present invention, for each candidate computing device, the available amount of multiple resources can be evaluated. For CPU resources, it is necessary to consider the current CPU utilization rate. The available CPU resources can be obtained by subtracting the current CPU utilization rate from 100%. For example, if the CPU utilization rate of a server is 30%, then its available CPU resources are 70%.

[0106] For memory, the available memory is obtained by subtracting the currently used memory from the total memory. For example, if a server has a total memory of 32 GB and 10 GB is currently in use, its available memory is 22 GB.

[0107] For storage resources, the available storage can be calculated based on the total storage capacity and the used capacity. At the same time, for some special hardware resources, such as GPUs, it is necessary to evaluate the available amount of video memory and the number of idle computing cores, etc.

[0108] For the available amount of network bandwidth, the currently used network bandwidth can be determined through network monitoring tools or the built-in network monitoring function of the system, and the available bandwidth is obtained by subtracting the used bandwidth from the total bandwidth.

[0109] A comprehensive evaluation method can be adopted to determine the available resource amount of each computing power device. For example, weights are assigned to different resources, the available resource amount of the CPU is multiplied by the CPU weight, the available resource amount of the memory is multiplied by the memory weight, etc., to calculate the comprehensive available resource score of each device. After sorting according to the scores, according to the sorting result, directly select the last device or the first device in the sorting result as the target computing power device. For example, in the sorting result from high to low, the first device is the computing power device with the largest available resource amount, and it is determined as the target computing power device.

[0110] In an embodiment of the present invention, according to the computing task and the device information of the computing power device, the initial computing power device is determined, including: constructing a computing power resource pool according to the device information of the computing power device, where the computing power resource pool includes the device information corresponding to different computing power devices; querying the initial computing power device corresponding to the computing task in the computing power resource pool.

[0111] In the embodiment of the present invention, the detailed device information of each computing power device can be obtained through means such as a V2V switch. As mentioned above, this information covers hardware information (such as the number of CPU cores, memory size, storage capacity, GPU performance, etc.), software information (such as operating system, installed software, network configuration, etc.), and the current running state (such as CPU usage rate, memory occupancy rate, etc.).

[0112] Then select a suitable data structure to store this device information. For example, a database table structure can be used, where each row record represents the information of a computing power device, and the columns correspond to different information fields, such as device ID, number of CPU cores, memory size, etc. Data structures such as tree structures and hash tables can also be used. Which one to choose specifically can be determined according to actual requirements and the requirements of query efficiency; for example, if it is often necessary to quickly locate device information according to the device ID, a hash table may be a better choice; if it is necessary to organize device information according to a certain hierarchical relationship, a tree structure would be more appropriate.

[0113] The information of each computing power device collected can be filled into the computing power resource pool according to the designed data structure. For example, the information such as the number of CPU cores of a certain computing power device being 8 and the memory size being 16GB can be correspondingly written into the corresponding fields of the database table to complete the storage of the device information in the resource pool.

[0114] After obtaining the computing task, various requirements of the computing task can be parsed, including the computing type (such as scientific computing, data processing, graphics rendering, etc.), the required resources (such as the required number of CPU cores, memory size, storage capacity, GPU performance, etc.), the task priority, the data input and output requirements, etc. For example, a deep learning training task may require a large amount of GPU computing resources and also has certain requirements for memory and storage.

[0115] Then, according to the requirements of the computing task, the query conditions are determined. For example, if the computing task requires at least 4 CPU cores and 8GB of memory, then the query condition is to filter out the computing power devices in the computing power resource pool with the number of CPU cores greater than or equal to 4 and the memory size greater than or equal to 8GB, and then use the corresponding query language or tool to perform a query operation in the computing power resource pool to obtain the corresponding initial computing power devices.

[0116] In an embodiment of the present invention, according to the operating state of the initial computing power device, candidate computing power devices are determined, including: determining the computing power devices with an idle operating state in the initial computing power devices as candidate computing power devices.

[0117] In the embodiment of the present invention, a running state query request can be sent to the initial computing power device through a V2V switch. After receiving the request, the initial computing power device feeds back its own running state information to the V2V switch, and then the V2V switch forwards it to the corresponding scheduling platform. The running state information includes but is not limited to key indicators such as the current CPU usage rate, memory occupancy rate, and whether other tasks are being processed on the device. Based on these indicators, it is determined whether the device is in an idle state.

[0118] For each initial computing power device, the management system makes a judgment based on the pre-set idle state judgment criteria. For example, if it is set that the CPU usage rate is lower than 10%, the memory occupancy rate is lower than 20%, and there are no other computing tasks being executed, then the device is considered to be in an idle state. Different systems may adjust these judgment criteria according to the actual situation and business requirements. For example, for some computing tasks that are more sensitive to resources, the CPU usage rate standard for the idle state may be set lower, such as 5%.

[0119] Select all the initial computing power devices that meet the idle state judgment criteria and determine them as candidate computing power devices. These candidate computing power devices will enter the next screening process to finally determine the target computing power device for executing the computing task. For example, after judgment, 3 devices among the initial computing power devices meet the idle state criteria, then these 3 devices become candidate computing power devices.

[0120] Such as Figure 3 , which shows the flowchart of a V2V-based intelligent computing virtualized computing power allocation method provided by an embodiment of the present invention. Each computing power server registers computing power resources to the V2V network management system through a Vision Network V2V switch, and the V2V network management system constructs a resource pool according to the computing power resources; each computing power server is communicatively connected through a VRB switch. After the computing power scheduling system receives a user's computing task, it analyzes the computing task to obtain resource requirement parameters, and then, according to the resource requirement parameters determined in the task analysis stage, quickly searches and filters in the registered computing power resource pool to find the initial computing power servers, and then obtains the operating states of each initial computing power server. According to the operating states of each initial computing power server, a candidate computing power server table is determined, and then the target computing power server is determined in the candidate computing power server table. After determining the target computing power server, the computing power scheduling system distributes the computing task to the corresponding target computing power server through the Vision Network V2V switch.

[0121] The present invention discloses a V2V-based intelligent computing virtualized computing power allocation method. The present invention can use a VRB switch to transmit data involved in the process of a target computing power device executing a computing task based on the VRB protocol. The VRB protocol is usually designed to meet the requirements of efficient data transmission under a specific network architecture, and can ensure the rapid and accurate transmission of data between computing power devices, reduce data transmission latency and packet loss rate, and greatly improve the data transmission efficiency and reliability between computing power resources; since the VRB protocol communicates based on the Vision Network protocol, relying on the communication node exchange mechanism of the V2V protocol can optimize the data flow path, and ensure efficient cooperation of computing power resources and stable data interaction from the communication basic level, reducing the cumbersome processes brought by intermediate nodes in the traditional multi-layer addressing method; the present invention can effectively solve the problems existing in the data transmission process of computing power resources in the traditional network architecture, and remove obstacles for the data transmission between computing power resources; it can dynamically determine the target computing power device according to the actual task requirements, realize the rapid and accurate positioning of the target computing power device, and improve the efficiency of establishing a communication link.

[0122] It should be noted that for the method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the embodiments of the present invention are not limited by the described action sequences, because according to the embodiments of the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0123] Referring to Figure 4 , a structural block diagram of a computing power allocation structure 20 provided by an embodiment of the present invention based on a VRB network architecture is shown. The structure includes a task scheduling platform 201 and a VRB network architecture 202. The VRB network architecture includes a plurality of computing power devices 2021 and a VRB switch 2022. The computing power devices are connected through the VRB switch.

[0124] The task scheduling platform 201 is used to obtain the device information of a plurality of computing power devices 2021 in the VRB network architecture; obtain a computing task, and determine a target computing power device according to the computing task and the device information of the computing power device; schedule the target computing power device to execute the computing task.

[0125] The VRB switch 2022 is used to transmit the data involved in the process of the target computing power device executing the computing task based on the VRB protocol.

[0126] In an embodiment of the present invention, a plurality of computing power devices are deployed in a plurality of clusters. The computing power devices in the same cluster are connected through the same VRB switch; the computing power devices between different clusters are connected through the VRB switches corresponding to different clusters.

[0127] The VRB switch corresponding to the same cluster is used to transmit the data involved in the process of the target computing power device in the same cluster executing the computing task based on the VRB protocol.

[0128] The VRB switch corresponding to different clusters is used to transmit the data involved in the process of the target computing power device in different clusters executing the computing task based on the VRB protocol.

[0129] In an embodiment of the present invention, a plurality of computing power devices in the VRB network architecture are respectively connected to a V2V switch. The task scheduling platform is used to determine an initial computing power device according to the computing task and the device information of the computing power device.

[0130] The V2V switch is used to obtain the operating status of the initial computing power device and send it to the task scheduling platform.

[0131] The task scheduling platform is used to determine candidate computing power devices according to the operating status of the initial computing power device; determine the target computing power device among the candidate computing power devices.

[0132] In an embodiment of the present invention, a V2V switch is configured to obtain device information of each computing power device in a VRB network architecture and send it to a task scheduling platform.

[0133] In an embodiment of the present invention, a computing task includes multiple subtasks and the priorities of each subtask among the multiple subtasks;

[0134] The task scheduling platform is configured to determine the order in which each subtask is executed by a target computing power device corresponding to each subtask according to the priorities of each subtask;

[0135] The V2V switch is configured to send multiple subtasks to the target computing power device corresponding to each subtask in sequence.

[0136] In an embodiment of the present invention, the task scheduling platform is configured to determine the available resource amount of each computing power device among candidate computing power devices; sort each computing power device according to the available resource amount of each computing power device to obtain a sorting result; and determine the computing power device with the largest available resource amount in the sorting result as the target computing power device.

[0137] In an embodiment of the present invention, the task scheduling platform is configured to construct a computing power resource pool according to the device information of the computing power device, where the computing power resource pool includes device information corresponding to different computing power devices; and query an initial computing power device corresponding to a computing task in the computing power resource pool.

[0138] In an embodiment of the present invention, the task scheduling platform is configured to determine a computing power device with an idle running state among the initial computing power devices as a candidate computing power device.

[0139] The present invention discloses a computing power allocation structure based on a VRB network architecture. The present invention can use a VRB switch to transmit data involved in the process of a target computing power device executing a computing task based on the VRB protocol. The VRB protocol is usually designed to meet the requirements of efficient data transmission under a specific network architecture, which can ensure the fast and accurate transmission of data between computing power devices, reduce data transmission delay and packet loss rate, and greatly improve the data transmission efficiency and reliability between computing power resources; since the VRB protocol communicates based on the visual networking protocol, the communication node exchange mechanism relying on the V2V protocol can optimize the data flow path, which ensures the efficient collaboration of computing power resources and stable data interaction from the communication basic level, and reduces the cumbersome processes brought by intermediate nodes under the traditional multi-layer addressing method; the present invention can effectively solve the problems existing in the data transmission process of computing power resources in the traditional network architecture and remove obstacles for the data transmission between computing power resources; it can dynamically determine the target computing power device according to the actual task requirements, realize the rapid and accurate positioning of the target computing power device, and improve the efficiency of establishing a communication link.

[0140] Reference Figure 5 , which shows a structural block diagram of an intelligent computing virtualized computing power allocation device provided by an embodiment of the present invention, applied to a VRB network architecture. The VRB network architecture includes multiple computing power devices and a VRB switch, and the computing power devices are connected through the VRB switch; the device may include the following modules:

[0141] The first acquisition module 301 is used to acquire the device information of multiple computing power devices in the VRB network architecture;

[0142] The second acquisition module 302 is used to acquire a computing task, and determine a target computing power device according to the computing task and the device information of the computing power device;

[0143] The scheduling module 303 is used to schedule the target computing power device to execute the computing task, and transmit the data involved in the process of the target computing power device executing the computing task through the VRB switch based on the VRB protocol.

[0144] The present invention discloses an intelligent computing virtualized computing power allocation device based on V2V. The present invention can use a VRB switch to transmit the data involved in the process of a target computing power device executing a computing task based on the VRB protocol. The VRB protocol is usually designed to meet the requirements of efficient data transmission under a specific network architecture, and can ensure the rapid and accurate transmission of data between computing power devices, reduce data transmission latency and packet loss rate, and greatly improve the data transmission efficiency and reliability between computing power resources; since the VRB protocol communicates based on the visual networking protocol, the communication node exchange mechanism relying on the V2V protocol can optimize the data flow path, guaranteeing the efficient cooperation of computing power resources and stable data interaction from the communication basic level, and reducing the cumbersome processes brought by intermediate nodes under the traditional multi-layer addressing method; the present invention can effectively solve the problems existing in the data transmission process of computing power resources in the traditional network architecture, removing obstacles for the data transmission between computing power resources; it can dynamically determine the target computing power device according to the actual task requirements, realize the rapid and accurate positioning of the target computing power device, and improve the efficiency of establishing a communication link.

[0145] In an embodiment of the present invention, multiple computing power devices are deployed in multiple clusters, and the computing power devices in the same cluster are connected through the same VRB switch; the computing power devices between different clusters are connected through the VRB switches corresponding to different clusters;

[0146] The scheduling module 303 includes:

[0147] The first transmission sub-module is used to transmit, through the VRB switch corresponding to the cluster, the data involved in the process of the target computing power device in the same cluster executing the computing task based on the VRB protocol;

[0148] The second transmission sub-module is used to transmit data involved in the process of the target computing devices in different clusters executing computing tasks through the VRB switches corresponding to different clusters based on the VRB protocol.

[0149] In an embodiment of the present invention, multiple computing devices in the VRB network architecture are respectively connected to the V2V switch. The second acquisition module 302 may include:

[0150] The first determination sub-module is used to determine the initial computing device according to the computing task and the device information of the computing device;

[0151] The first acquisition sub-module is used to obtain the operating status of the initial computing device through the V2V switch;

[0152] The second determination sub-module is used to determine the candidate computing devices according to the operating status of the initial computing device;

[0153] The third determination sub-module is used to determine the target computing device among the candidate computing devices.

[0154] In an embodiment of the present invention, the first acquisition module 301 may include

[0155] The second acquisition sub-module is used to obtain the device information of each computing device in the VRB network architecture through the V2V switch

[0156] In an embodiment of the present invention, the first acquisition sub-module includes:

[0157] The sending unit is used to send a query request to the initial computing device through the V2V switch, and the query request is used to obtain the operating status;

[0158] The receiving unit is used to receive the operating status sent by the initial computing device through the V2V switch.

[0159] In an embodiment of the present invention, the computing task includes multiple sub-tasks and the priorities of each sub-task among the multiple sub-tasks;

[0160] The scheduling module 303 includes:

[0161] The fourth determination sub-module is used to determine the order of the target computing devices corresponding to each sub-task to execute the tasks according to the priorities of each sub-task;

[0162] The sending sub-module is used to send multiple sub-tasks to the target computing devices corresponding to each sub-task in sequence through the V2V switch.

[0163] In an embodiment of the present invention, the third determination sub-module may include:

[0164] A first determination unit, configured to determine the available resource amount of each computing power device in candidate computing power devices;

[0165] A sorting unit, configured to sort each computing power device according to the available resource amount of each computing power device to obtain a sorting result;

[0166] A second determination unit, configured to determine the computing power device with the largest available resource amount in the sorting result as the target computing power device.

[0167] In an embodiment of the present invention, the first determination sub-module may include:

[0168] A construction unit, configured to construct a computing power resource pool according to the device information of the computing power device, where the computing power resource pool includes the device information corresponding to different computing power devices;

[0169] A query unit, configured to query the initial computing power device corresponding to the computing task in the computing power resource pool.

[0170] In an embodiment of the present invention, the second determination sub-module may include:

[0171] A third determination unit, configured to determine the computing power device with an idle operating state in the initial computing power device as a candidate computing power device.

[0172] The present invention discloses an intelligent computing virtualized computing power allocation device based on V2V. The present invention can use a VRB switch to transmit data involved in the process of a target computing power device executing a computing task based on the VRB protocol. The VRB protocol is usually designed to meet the requirements of efficient data transmission under a specific network architecture, which can ensure the rapid and accurate transmission of data between computing power devices, reduce data transmission delay and packet loss rate, and greatly improve the data transmission efficiency and reliability between computing power resources; since the VRB protocol communicates based on the visual networking protocol, the communication node switching mechanism relying on the V2V protocol can optimize the data flow path, ensuring efficient cooperation of computing power resources and stable data interaction from the communication basic level, and reducing the cumbersome processes brought by intermediate nodes under the traditional multi-layer addressing method; the present invention can effectively solve the problems existing in the data transmission process of computing power resources in the traditional network architecture, removing obstacles for the data transmission between computing power resources; it can dynamically determine the target computing power device according to the actual task requirements, realize the rapid and accurate positioning of the target computing power device, and improve the efficiency of establishing a communication link.

[0173] For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and for the related parts, please refer to the partial description of the method embodiment.

[0174] An embodiment of the present invention further provides an electronic device, including:

[0175] It includes a processor, a memory, and a computer program stored on the memory and capable of running on the processor. When the computer program is executed by the processor, it implements each process of the above-mentioned embodiment of the intelligent computing virtualized computing power allocation method based on V2V, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0176] The embodiment of the present invention also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by the processor, it implements each process of the above-mentioned embodiment of the intelligent computing virtualized computing power allocation method based on V2V, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0177] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other.

[0178] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a device, or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0179] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowcharts and / or block diagrams can also be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing terminal devices to generate a machine, so that the instructions executed by the processors of the computer or other programmable data processing terminal devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0180] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions in Figure 1 one process or multiple processes and / or blocks Figure 1The functions specified in one or more boxes.

[0181] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal device provide for implementing steps of the functions specified in one Figure 1 one process or multiple processes and / or boxes Figure 1 steps of the functions specified in one or more boxes.

[0182] Although the preferred embodiments of the embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

[0183] Finally, it should also 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 such 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 terminal device comprising 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 terminal device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, method, article or terminal device comprising the element.

[0184] The above has introduced in detail a V2V-based intelligent computing virtualized computing power allocation method, structure, device, equipment and storage 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 V2V-based intelligent computing virtualized computing power allocation method, characterized in that, Applied to a VRB network architecture, the VRB network architecture includes multiple computing power devices and a VRB switch, and the computing power devices are connected through the VRB switch; the method includes: Obtain the device information of multiple computing power devices in the VRB network architecture; Obtain a computing task, and determine a target computing power device according to the computing task and the device information of the computing power device; Schedule the target computing power device to execute the computing task, and transmit the data involved in the process of the target computing power device executing the computing task based on the VRB protocol through the VRB switch.

2. The method according to claim 1, characterized in that, The multiple computing power devices are deployed in multiple clusters, and the computing power devices in the same cluster are connected through the same VRB switch; the computing power devices between different clusters are connected through the VRB switches corresponding to different clusters; The transmitting the data involved in the process of the target computing power device executing the computing task based on the VRB protocol through the VRB switch includes: Transmit the data involved in the process of the target computing power device in the same cluster executing the computing task based on the VRB protocol through the VRB switch corresponding to the cluster; Transmit the data involved in the process of the target computing power device in different clusters executing the computing task based on the VRB protocol through the VRB switches corresponding to different clusters.

3. The method according to claim 1, wherein Multiple computing power devices in the VRB network architecture are respectively connected to a V2V switch, and determining a target computing power device according to the computing task and the device information of the computing power device includes: Determine an initial computing power device according to the computing task and the device information of the computing power device; Obtain the operating status of the initial computing power device through the V2V switch; Determine candidate computing power devices according to the operating status of the initial computing power device; Determine a target computing power device among the candidate computing power devices.

4. The method according to claim 3, wherein The obtaining the device information of the computing power device in the VRB network architecture includes: Obtain the device information of each computing power device in the VRB network architecture through the V2V switch.

5. The method according to claim 3, characterized in that, The obtaining the operating status of the initial computing power device through the V2V switch includes: Send a query request to the initial computing power device through the V2V switch, and the query request is used to obtain the operating status; Receive the operating status sent by the initial computing power device through the V2V switch.

6. The method according to claim 3, characterized in that, The computing task includes multiple subtasks and the priority of each subtask among the multiple subtasks; The scheduling the target computing power device to execute the computing task includes: Determine the order of the target computing power devices corresponding to the respective subtasks to execute the tasks according to the priorities of the respective subtasks; Send the multiple subtasks to the target computing power devices corresponding to the respective subtasks in the order through the V2V switch.

7. The method according to claim 3, wherein The determining a target computing power device among the candidate computing power devices includes: Determine the available resource amount of each computing power device among the candidate computing power devices; Sort the respective computing power devices according to the available resource amount of each computing power device to obtain a sorting result; Determine the computing power device with the largest available resource amount in the sorting result as the target computing power device.

8. The method according to claim 3, wherein The determining of the initial computing power device according to the computing task and the device information of the computing power device includes: Construct a computing power resource pool according to the device information of the computing power device, where the computing power resource pool includes the device information corresponding to different computing power devices; Query the initial computing power device corresponding to the computing task in the computing power resource pool.

9. The method according to claim 3, wherein The determining of the candidate computing power device according to the running state of the initial computing power device includes: Determine the computing power devices with an idle running state in the initial computing power devices as candidate computing power devices.

10. A V2V-based intelligent computing virtualized computing power allocation structure, characterized in that, The structure includes a task scheduling platform and a VRB network architecture. The VRB network architecture includes multiple computing power devices and VRB switches, and the computing power devices are connected through the VRB switches; The task scheduling platform is used to obtain the device information of multiple computing power devices in the VRB network architecture; Obtain a computing task, and determine the target computing power device according to the computing task and the device information of the computing power device; Schedule the target computing power device to execute the computing task; The VRB switch is used to transmit the data involved in the process of the target computing power device executing the computing task based on the VRB protocol.

11. The structure according to claim 10, wherein The multiple computing power devices are deployed in multiple clusters. The computing power devices in the same cluster are connected through the same VRB switch; the computing power devices between different clusters are connected through the VRB switches corresponding to different clusters; The VRB switch corresponding to the same cluster is used to transmit the data involved in the process of the target computing power device in the same cluster executing the computing task based on the VRB protocol; The VRB switch corresponding to different clusters is used to transmit the data involved in the process of the target computing power device in different clusters executing the computing task based on the VRB protocol.

12. The structure according to claim 10, wherein Multiple computing power devices in the VRB network architecture are respectively connected to the V2V switch. The task scheduling platform is used to determine the initial computing power device according to the computing task and the device information of the computing power device; The V2V switch is used to obtain the running state of the initial computing power device and send it to the task scheduling platform; The task scheduling platform is used to determine the candidate computing power device according to the running state of the initial computing power device; determine the target computing power device among the candidate computing power devices.

13. The structure according to claim 12, wherein, The V2V switch is used to obtain the device information of each computing power device in the VRB network architecture and send it to the task scheduling platform.

14. The structure according to claim 12, wherein, The computing task includes multiple subtasks and the priority of each subtask among the multiple subtasks; The task scheduling platform is used to determine the order of the target computing power devices corresponding to each subtask to execute the tasks according to the priority of each subtask; The V2V switch is used to send the multiple subtasks to the target computing power devices corresponding to each subtask in accordance with the order.

15. The structure according to claim 12, characterized in that, The task scheduling platform is used to determine the available resource amounts of each computing power device among the candidate computing power devices; sort each of the computing power devices according to the available resource amounts of the computing power devices to obtain a sorting result; and determine the computing power device with the largest available resource amount in the sorting result as the target computing power device.

16. The structure according to claim 12, wherein, The task scheduling platform is used to construct a computing power resource pool according to the device information of the computing power devices, where the computing power resource pool includes the device information corresponding to different computing power devices; and query the initial computing power device corresponding to the computing task in the computing power resource pool.

17. The structure according to claim 12, characterized in that, The task scheduling platform is used to determine the computing power devices with an idle operating state among the initial computing power devices as candidate computing power devices.

18. A V2V-based intelligent computing virtualized computing power allocation device, applied to the VRB network architecture, characterized in that The VRB network architecture includes multiple computing power devices and VRB switches, and the computing power devices are connected through the VRB switches; the device includes: A first acquisition module, configured to acquire the device information of multiple computing power devices in the VRB network architecture; A second acquisition module, configured to acquire a computing task, and determine a target computing power device according to the computing task and the device information of the computing power devices; A scheduling module, configured to schedule the target computing power device to execute the computing task, and transmit the data involved in the process of the target computing power device executing the computing task through the VRB switch based on the VRB protocol.

19. An electronic device, characterized in that, including: A processor, a memory, and a computer program stored on the memory and capable of running on the processor, where when the computer program is executed by the processor, the steps of the V2V-based intelligent computing virtualized computing power allocation method according to any one of claims 1-9 are implemented.

20. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, the steps of the V2V-based intelligent computing virtualized computing power allocation method according to any one of claims 1-9 are implemented.

Images