A method, device, terminal and storage medium for edge-end collaborative intelligent resource scheduling
Through device virtualization technology and scheduling strategies, the resource management problems of heterogeneous edge-end devices in deep learning tasks are solved, efficient unified abstraction and automated scheduling of resources are realized, and the efficiency and rationality of edge-end devices collaborative work are improved.
Patent Information
- Application Number
- CN202210270214.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-18
- Publication Date
- 2024-12-17
- Estimated Expiration
- 2042-03-18
AI Technical Summary
The existing technology is difficult to effectively integrate and collaborate in managing the AI capabilities of multiple heterogeneous edge devices, especially in deep learning tasks, which is difficult to achieve reasonable allocation and management of resources, resulting in low utilization efficiency, poor fairness and response time of computing resources. long.
Through device virtualization technology, multi-heterogeneous edge devices are virtualized into logical units, and scheduling strategies for multi-heterogeneous resource management are designed, device resource conditions are collected, pre-trained models and jobs are issued, tasks are executed and results are feedback. Realize refined management and scheduling of resources.
It realizes efficient unified abstraction and automated scheduling of heterogeneous resources, simplifies equipment management, improves the efficiency and rationality of edge-end equipment collaborative work, and enables the model to automatically learn appropriate parameters and configurations without manual intervention.
Smart Images
Figure CN114584563B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of edge device resource scheduling, and in particular to an edge collaborative intelligent resource scheduling method, device, terminal and storage medium. Background Art
[0002] In order to meet the needs of diversified computing, more and more scenarios have begun to introduce hardware such as NPU, GPU, FPGA for acceleration, especially in the field of artificial intelligence. At present, heterogeneous computing represented by NPU has become a new generation of computing architecture to accelerate AI innovation. The number of smart devices owned by users is increasing, and the scenarios that require AI functions are constantly enriching. In this context, the demand for collaborative AI capabilities between multiple devices is gradually increasing. How to integrate these scattered AI capabilities distributed in various devices, and find the best device combination to provide services to users according to different scenarios and user needs is a problem that needs to be explored now.
[0003] In order to ensure the effective utilization, fairness, mutual exclusion management, fast response, and queue maintenance of computing resources under full load when multiple heterogeneous resources are running in various task scenarios, a set of efficient and sophisticated allocation principles and scheduling management mechanisms are required. Figure 1 The figure shows a schematic diagram of the collaborative topology of edge devices. In order to give full play to the computing power advantages of edge servers (large computing power, high latency, and strong dependence on communication networks) and embedded intelligent terminals (low latency, small computing power, and weak dependence on communication networks), it is necessary to consider the collaborative design between edge devices so that embedded edge devices can better manage embedded intelligent terminals and achieve unified scheduling of resources for the two types of devices. However, it is difficult to uniformly and reasonably schedule computing resources with different types of instruction sets and different architectures, and it is difficult to reasonably allocate and manage resources for different deep learning tasks. This phenomenon becomes more prominent as the scale and complexity of computing increase. Summary of the invention
[0004] To solve the above problems, the present invention provides an edge-to-edge collaborative intelligent resource scheduling method, device, terminal and storage medium, which solves the problem of abstracting multiple heterogeneous resources through device virtualization technology, solves the refined allocation and management of multiple heterogeneous resources through scheduling strategy design technology for multi-heterogeneous resource management, and realizes an efficient scheduling mechanism to perform refined management of heterogeneous resources used in deep learning tasks.
[0005] In a first aspect, the technical solution of the present invention provides an edge collaborative intelligent resource scheduling method, comprising the following steps:
[0006] Virtualize multiple heterogeneous edge devices;
[0007] Collect and count the resource conditions of edge devices;
[0008] According to the resource conditions of the edge device, the cloud pre-trained model is sent to the corresponding edge device;
[0009] Allocate jobs to corresponding edge devices based on job resource requirements and the model required for the job;
[0010] Call the corresponding model on the assigned edge device to execute the job;
[0011] Feedback the job execution results to the cloud or resource scheduling platform.
[0012] Furthermore, virtualizing multiple heterogeneous edge devices is specifically as follows: virtualizing multiple heterogeneous edge devices into logical units of system size.
[0013] Further, the multi-heterogeneous edge devices include an edge server and an embedded intelligent terminal, and the edge server communicates with the plurality of embedded intelligent terminals;
[0014] The resource conditions of edge devices include the available resources, cost, quality, and latency of the computing power and bandwidth resources of the edge devices.
[0015] Furthermore, the job execution results are fed back to the cloud or resource scheduling platform, specifically:
[0016] Determine whether the task is completed;
[0017] If completed, the result will be fed back to the cloud; if not completed, the result will be fed back to the resource scheduling platform, which will then assign the result to the corresponding edge device as the next job.
[0018] In a second aspect, the technical solution of the present invention provides an edge collaborative intelligent resource scheduling device, comprising:
[0019] Virtualization module: virtualizes multiple heterogeneous edge devices;
[0020] Resource condition collection and statistics module: collects and counts the resource conditions of edge devices;
[0021] Model delivery module: delivers the cloud-based pre-trained model to the corresponding edge device based on the resource conditions of the edge device;
[0022] Job distribution module: allocates jobs to corresponding edge devices based on job resource requirements and the model required for the job;
[0023] Job execution module: calls the corresponding model on the assigned edge device to execute the job;
[0024] Result feedback module: Feedback the job execution results to the cloud or resource scheduling platform.
[0025] Furthermore, the virtualization module virtualizes the multiple heterogeneous edge devices in the following ways: virtualizing the multiple heterogeneous edge devices into logical units of the system size.
[0026] Further, the multi-heterogeneous edge devices include an edge server and an embedded intelligent terminal, and the edge server communicates with the plurality of embedded intelligent terminals;
[0027] The resource conditions of edge devices include the available resources, cost, quality, and latency of the computing power and bandwidth resources of the edge devices.
[0028] Furthermore, the result feedback module feeds back the job execution results to the cloud or resource scheduling platform, specifically:
[0029] Determine whether the task is completed;
[0030] If completed, the result will be fed back to the cloud; if not completed, the result will be fed back to the resource scheduling platform, which will then assign the result to the corresponding edge device as the next job.
[0031] In a third aspect, the technical solution of the present invention provides a terminal, including:
[0032] A memory, used to store edge collaborative intelligent resource scheduling programs;
[0033] A processor is used to implement the steps of the edge collaborative intelligent resource scheduling method as described in any of the above items when executing the edge collaborative intelligent resource scheduling program.
[0034] In a fourth aspect, the technical solution of the present invention provides a computer-readable storage medium, on which a edge-side collaborative intelligent resource scheduling program is stored. When the edge-side collaborative intelligent resource scheduling program is executed by a processor, the steps of the edge-side collaborative intelligent resource scheduling method as described in any one of the above items are implemented.
[0035] Compared with the prior art, the edge collaborative intelligent resource scheduling method, device, terminal and storage medium provided by the present invention have the following beneficial effects: through device virtualization technology, edge devices are virtualized, and the virtualized devices are transmitted to the container for use, so as to realize the unified abstraction of various resources. Combined with the intelligent autonomous scheduling strategy, the model can automatically learn the appropriate parameters and configurations without human intervention, which simplifies device management and user use, and greatly improves the efficiency and rationality of the collaborative work of edge devices. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the embodiments of the present application or the technical solutions of the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0037] Figure 1 Schematic diagram of the collaborative topology structure of edge devices.
[0038] Figure 2 A schematic diagram of the principles of an edge collaborative intelligent resource scheduling method provided in Example 1 of the present invention.
[0039] Figure 3 A flowchart of an edge collaborative intelligent resource scheduling method provided in Example 1 of the present invention.
[0040] Figure 4 A schematic block diagram of the structure of an edge collaborative intelligent resource scheduling device provided in Example 2 of the present invention.
[0041] Figure 5 A schematic diagram of the structure of a terminal provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION
[0042] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.
[0043] Embodiment 1
[0044] This embodiment 1 provides an edge-end collaborative intelligent resource scheduling method, such as Figure 2 The figure shows a schematic diagram of the principle flow of this method. The resource scheduling platform is intelligent. It first virtualizes the edge devices, then unifies the resource abstraction, reports the resource monitoring, allocates the device resources, and executes the job. If the task is completed, the result is fed back to the cloud. If the task is not completed and the job needs to be sent to the next-level device, the result is fed back to the resource scheduling platform to reallocate resources.
[0045] like Figure 3 As shown, the edge collaborative intelligent resource scheduling method provided in this embodiment 1 specifically includes the following steps.
[0046] S101, virtualize multiple heterogeneous edge devices.
[0047] It should be noted that the multi-heterogeneous edge devices include an edge server 1 and an embedded intelligent terminal 2, and the edge server 1 communicates with multiple embedded intelligent terminals 2.
[0048] This embodiment virtualizes multiple heterogeneous edge devices by virtualizing multiple heterogeneous resources into logical units of the system size, so that multiple containers can share the computing power of the system.
[0049] S102, collecting and counting resource conditions of edge devices.
[0050] This embodiment counts the resource conditions of the edge devices for subsequent training models and job delivery. Specifically, the resource conditions of the edge devices collected and counted in this embodiment include the available resources, cost, quality and latency of the computing power resources and bandwidth resources of the edge devices.
[0051] S103: According to the resource conditions of the edge device, the cloud pre-trained model is sent to the corresponding edge device.
[0052] It should be noted that this embodiment reports the resource conditions of each edge device in the local area network, and the resource scheduling platform sends the pre-trained model in the cloud to different edge devices based on the collected resource conditions.
[0053] S104: Allocate the job to the corresponding edge device according to the job resource requirements and the model required by the job.
[0054] It should be noted that, in this embodiment, the job is assigned to the edge server or the embedded intelligent terminal, such as the embedded intelligent terminal performs rough classification and cooperates with the edge server to perform further fine identification.
[0055] S105: Call the corresponding model on the allocated edge device to execute the job.
[0056] S106, feeding back the job execution result to the cloud or resource scheduling platform.
[0057] After a job is executed once, it may complete the task or it may not complete the task and need to be continued. Therefore, we first determine whether the task is completed. If it is completed, the result is fed back to the cloud. If it is not completed, the result is fed back to the resource scheduling platform, and the resource scheduling platform allocates the result to the corresponding edge device as the next job.
[0058] It should be noted that during the execution of the job, the resource scheduling platform continuously synchronizes data status with other edge devices in the local area network.
[0059] The edge collaborative intelligent resource scheduling method provided in this embodiment virtualizes the edge devices through device virtualization technology, and transmits the virtualized devices to the container for use, thereby realizing the unified abstraction of various resources. Combined with the intelligent autonomous scheduling strategy, the model can automatically learn the appropriate parameters and configurations without human intervention, which simplifies device management and user use, and greatly improves the efficiency and rationality of the collaborative work of edge devices.
[0060] Embodiment 2
[0061] Based on the edge collaborative intelligent resource scheduling method provided in Example 1, this embodiment provides an edge collaborative intelligent resource scheduling device for implementing the aforementioned method.
[0062] like Figure 4 As shown, the device provided in this embodiment is configured on a resource scheduling platform and includes the following functional modules.
[0063] Virtualization module 101: virtualizes multiple heterogeneous edge devices.
[0064] Specifically, multiple heterogeneous edge devices are virtualized into system-sized logical units so that multiple containers can share the computing power of the system.
[0065] It should be noted that the multi-heterogeneous edge devices include an edge server 1 and an embedded intelligent terminal 2, and the edge server 1 communicates with multiple embedded intelligent terminals 2.
[0066] Resource condition collection and statistics module 102: collects and counts resource conditions of edge devices.
[0067] The resource conditions of edge devices include the available resources, cost, quality, and latency of the computing power and bandwidth resources of the edge devices.
[0068] Model sending module 103: sends the cloud pre-trained model to the corresponding edge device according to the resource conditions of the edge device.
[0069] Job dispatching module 104: allocates the job to the corresponding edge device according to the job resource requirements and the model required by the job.
[0070] Job execution module 105: calls the corresponding model on the allocated edge device to execute the job.
[0071] Result feedback module 106: Feedback the job execution result to the cloud or resource scheduling platform.
[0072] Specifically, the result feedback module 106 determines whether the task is completed. If it is completed, the result is fed back to the cloud. If it is not completed, the result is fed back to the resource scheduling platform, and the resource scheduling platform allocates the result to the corresponding edge device as the next job.
[0073] The edge collaborative intelligent resource scheduling device of this embodiment is used to implement the aforementioned edge collaborative intelligent resource scheduling method. Therefore, the specific implementation method of the device can be seen in the embodiment part of the edge collaborative intelligent resource scheduling method in the previous text. Therefore, its specific implementation method can refer to the description of the corresponding embodiments of each part, which will not be introduced in detail here.
[0074] In addition, since the edge collaborative intelligent resource scheduling device of this embodiment is used to implement the aforementioned edge collaborative intelligent resource scheduling method, its function corresponds to that of the aforementioned method and will not be repeated here.
[0075] Embodiment 3
[0076] Figure 5 A schematic diagram of the structure of a terminal device 500 provided in an embodiment of the present invention includes: a processor 510, a memory 520 and a communication unit 530. The processor 510 is used to implement the following steps when implementing the edge collaborative intelligent resource scheduling program stored in the memory 520:
[0077] Virtualize multiple heterogeneous edge devices;
[0078] Collect and count the resource conditions of edge devices;
[0079] According to the resource conditions of the edge device, the cloud pre-trained model is sent to the corresponding edge device;
[0080] Allocate jobs to corresponding edge devices based on job resource requirements and the model required for the job;
[0081] Call the corresponding model on the assigned edge device to execute the job;
[0082] Feedback the job execution results to the cloud or resource scheduling platform.
[0083] The present invention virtualizes edge devices through device virtualization technology, transmits the virtualized devices to the container for use, realizes the unified abstraction of various resources, and combines with intelligent autonomous scheduling strategies, so that the model can automatically learn the appropriate parameters and configurations without human intervention, simplifies device management and user use, and greatly improves the efficiency and rationality of the collaborative work of edge devices.
[0084] The terminal device 500 includes a processor 510, a memory 520 and a communication unit 530. These components communicate via one or more buses. It can be understood by those skilled in the art that the structure of the server shown in the figure does not constitute a limitation of the present invention, and it can be a bus structure or a star structure, and can also include more or less components than shown in the figure, or combine some components, or arrange the components differently.
[0085] The memory 520 can be used to store the execution instructions of the processor 510, and the memory 520 can be implemented by any type of volatile or non-volatile storage terminal or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. When the execution instructions in the memory 520 are executed by the processor 510, the terminal 500 can execute some or all of the steps in the following method embodiments.
[0086] The processor 510 is the control center of the storage terminal, and uses various interfaces and lines to connect various parts of the entire electronic terminal. It runs or executes software programs and / or modules stored in the memory 520, and calls data stored in the memory to perform various functions of the electronic terminal and / or process data. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of multiple packaged ICs with the same or different functions. For example, the processor 510 can only include a central processing unit (CPU). In the embodiment of the present invention, the CPU can be a single computing core or multiple computing cores.
[0087] The communication unit 530 is used to establish a communication channel so that the storage terminal can communicate with other terminals, receive user data sent by other terminals or send user data to other terminals.
[0088] Embodiment 4
[0089] The present invention also provides a computer storage medium, wherein the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).
[0090] The computer storage medium stores an edge collaborative intelligent resource scheduling program, and when the edge collaborative intelligent resource scheduling program is executed by a processor, the following steps are implemented:
[0091] Virtualize multiple heterogeneous edge devices;
[0092] Collect and count the resource conditions of edge devices;
[0093] According to the resource conditions of the edge device, the cloud pre-trained model is sent to the corresponding edge device;
[0094] Allocate jobs to corresponding edge devices based on job resource requirements and the model required for the job;
[0095] Call the corresponding model on the assigned edge device to execute the job;
[0096] Feedback the job execution results to the cloud or resource scheduling platform.
[0097] The present invention virtualizes edge devices through device virtualization technology, transmits the virtualized devices to the container for use, realizes the unified abstraction of various resources, and combines with intelligent autonomous scheduling strategies, so that the model can automatically learn the appropriate parameters and configurations without human intervention, simplifies device management and user use, and greatly improves the efficiency and rationality of the collaborative work of edge devices.
[0098] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention, in essence or in other words, the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes, including several instructions for enabling a computer terminal (which can be a personal computer, a server, or a second terminal, a network terminal, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention.
[0099] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0100] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0101] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0102] The above disclosure is only a preferred embodiment of the present invention, but the present invention is not limited thereto. Any non-creative changes that can be thought of by a person skilled in the art, as well as several improvements and modifications made without departing from the principle of the present invention, should fall within the protection scope of the present invention.
Claims
1. A method for edge collaborative intelligent resource scheduling, characterized in that: The following steps are involved: Virtualize multiple heterogeneous edge devices; Collect and count the resource conditions of edge devices; According to the resource conditions of the edge device, the cloud pre-trained model is sent to the corresponding edge device; Allocate jobs to corresponding edge devices based on job resource requirements and the model required for the job; Call the corresponding model on the assigned edge device to execute the job; Feedback the job execution results to the cloud or resource scheduling platform; Virtualizing multiple heterogeneous edge devices is as follows: Virtualize multiple heterogeneous edge devices into system-sized logical units, so that multiple containers can share the computing power of the system; The multi-heterogeneous edge devices include edge servers and embedded intelligent terminals. The edge server communicates with multiple embedded intelligent terminals. The resource conditions of edge devices include the available resources, cost, quality, and latency of edge device computing resources and bandwidth resources; Feedback the job execution results to the cloud or resource scheduling platform, specifically: Determine whether the task is completed; If the result is completed, the result will be fed back to the cloud. If the result is not completed, the result will be fed back to the resource scheduling platform, which will assign the result to the corresponding edge device as the next job. Virtualize the edge devices, then unify the resource abstraction, monitor and report the resources, allocate device resources, and execute the job. If the task is completed, the result will be fed back to the cloud. If the task is not completed and the job needs to be sent to the next level device, the result will be fed back to the resource scheduling platform to reallocate resources.
2. An edge-end collaborative intelligent resource scheduling device, characterized in that: include, Virtualization module: virtualizes multiple heterogeneous edge devices; Resource condition collection and statistics module: collects and counts the resource conditions of edge devices; Model delivery module: delivers the cloud-based pre-trained model to the corresponding edge device based on the resource conditions of the edge device; Job distribution module: allocates jobs to corresponding edge devices based on job resource requirements and the model required for the job; Job execution module: calls the corresponding model on the assigned edge device to execute the job; Result feedback module: Feedback the job execution results to the cloud or resource scheduling platform; The virtualization module virtualizes multiple heterogeneous edge devices as follows: Virtualize multiple heterogeneous edge devices into system-sized logical units; The multi-heterogeneous edge devices include edge servers and embedded intelligent terminals. The edge server communicates with multiple embedded intelligent terminals. The resource conditions of edge devices include the available resources, cost, quality, and latency of edge device computing resources and bandwidth resources; The result feedback module feeds back the job execution results to the cloud or resource scheduling platform, specifically: Determine whether the task is completed; If completed, the result will be fed back to the cloud; if not completed, the result will be fed back to the resource scheduling platform, which will then assign the result to the corresponding edge device as the next job.
3. A terminal, characterized in that: include: A memory, used to store edge collaborative intelligent resource scheduling programs; A processor is used to implement the steps of the edge collaborative intelligent resource scheduling method as described in claim 1 when executing the edge collaborative intelligent resource scheduling program.
4. A computer-readable storage medium, characterized in that: The readable storage medium stores an edge collaborative intelligent resource scheduling program, and when the edge collaborative intelligent resource scheduling program is executed by the processor, the steps of the edge collaborative intelligent resource scheduling method as claimed in claim 1 are implemented.