A resource scheduling method and an electronic device

Abstract

This application discloses a resource scheduling method and electronic device, relating to the field of computer technology. The method involves: acquiring a creation request sent by a target computing platform; selecting physical computing nodes from a target resource pool and sending the creation request to the physical computing nodes so that they can create virtual computing nodes; acquiring pending tasks sent by the target computing platform; acquiring the current status information of the physical computing nodes and allocating a first time slice to each physical computing node; allocating a second time slice to the virtual computing nodes after the first time slice is allocated based on the required time slice and current status information of the pending tasks; and sending the pending tasks to the virtual computing nodes after the second time slice is allocated so that the virtual computing nodes after the second time slice can be invoked to execute the pending tasks. This method addresses the problems of inflexible physical deployment, low resource scheduling efficiency, difficulty in integrating heterogeneous computing power, and limited transmission quality, thereby improving resource utilization efficiency and reducing bit error rate and transmission loss.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a resource scheduling method and an electronic device. Background Technology

[0002] In AI (Artificial Intelligence) large-scale model training and cloud computing scenarios, core problems arise such as low GPU (Graphics Processing Unit) resource utilization, insufficient scheduling efficiency, and physical limitations of traditional electrical interconnect architectures. Although existing GPU virtualization technologies can achieve single-machine resource partitioning, the transmission distance limitations of PCIe (Peripheral Component Interconnect Express) cables make it difficult to build elastic resource pools across physical devices. Furthermore, electrical signal transmission suffers from drawbacks such as high electromagnetic interference, high bit error rate, and significant power consumption.

[0003] It is evident that how to solve the problems of inflexible physical deployment, low resource scheduling efficiency, difficulty in integrating heterogeneous computing power, and limited transmission quality caused by the limitations of electrical signal transmission in traditional computing device resource pooling systems, and how to improve resource utilization efficiency, optimize transmission performance, and reduce bit error rate and transmission loss are problems that need to be solved by those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to provide a resource scheduling method and electronic device that can solve the problems of inflexible physical deployment, low resource scheduling efficiency, difficulty in integrating heterogeneous computing power, and limited transmission quality caused by limitations in electrical signal transmission in traditional computing device resource pooling systems. This improves resource utilization efficiency, optimizes transmission performance, and reduces bit error rate and transmission loss. The specific solution is as follows:

[0005] In a first aspect, this application discloses a resource scheduling method applied to a resource controller, the resource controller being connected to a target resource pool via optical interconnection, the target resource pool being configured with several physical computing nodes; wherein the method includes:

[0006] Obtain the creation request for creating a virtual computing node sent by the target computing platform;

[0007] Based on the creation request, the corresponding physical computing nodes are selected from the target resource pool, and the creation request is sent to the physical computing nodes so that the physical computing nodes can create the corresponding virtual computing nodes based on the creation request;

[0008] Forward the creation success response returned by the physical computing node to the target computing platform to obtain the pending tasks sent by the target computing platform;

[0009] Obtain the current status information of the physical computing nodes in the target resource pool, and allocate the corresponding first time slice to each physical computing node based on the current status information;

[0010] Based on the required time slice and current status information of the task to be processed, a second time slice is allocated to the virtual computing nodes in each physical computing node after the first time slice is allocated.

[0011] The task to be processed is sent to the virtual computing node after the second time slice is allocated, so that the virtual computing node after the second time slice is allocated can be invoked to execute the task to be processed.

[0012] Secondly, this application discloses an electronic device, including:

[0013] Memory, used to store computer programs;

[0014] A processor is used to implement the steps of the aforementioned resource scheduling method when executing a computer program.

[0015] Thirdly, this application discloses a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the aforementioned resource scheduling method.

[0016] As can be seen, this application provides a resource scheduling method, including: obtaining a creation request for creating a virtual computing node sent by a target computing platform; selecting the corresponding physical computing node from the target resource pool according to the creation request, and sending the creation request to the physical computing node so that the physical computing node can create the corresponding virtual computing node based on the creation request; forwarding the creation success response returned by the physical computing node to the target computing platform to obtain the pending task sent by the target computing platform; obtaining the current status information of the physical computing nodes in the target resource pool, and allocating a corresponding first time slice to each physical computing node based on the current status information; allocating a second time slice to the virtual computing nodes in each physical computing node after allocating the first time slice based on the required time slice of the pending task and the current status information; and sending the pending task to the virtual computing node after allocating the second time slice so that the virtual computing node after allocating the second time slice can be invoked to execute the pending task. This application applies to a resource controller, which is connected to a target resource pool via optical interconnect. The target resource pool contains several physical computing nodes. The controller receives creation requests for virtual computing nodes from a target computing platform, selects the corresponding physical computing nodes from the target resource pool based on the creation requests, and sends the creation requests to the physical computing nodes. These physical computing nodes are connected to the target resource pool via optical interconnect. Leveraging the long transmission distance and strong electromagnetic interference resistance of optical signals, this application overcomes the transmission distance limitations of traditional high-speed serial computer expansion bus standard cables, enabling flexible deployment of physical computing nodes within a range of hundreds of meters. The physical computing nodes create the corresponding virtual computing nodes based on the creation requests and forward the creation success response returned by the physical computing nodes to the target computing platform to obtain pending tasks sent by the target computing platform. This achieves physical decoupling, significantly reduces bit error rate and power loss, and ensures near-perfect performance during large-scale expansion. The system optimizes the performance of physical computing nodes, obtains the current status information of physical computing nodes in the target resource pool, allocates corresponding first time slices to each physical computing node based on the current status information, and allocates second time slices to virtual computing nodes in each physical computing node after the allocation of the first time slice based on the required time slice of the task to be processed and the current status information. Through time slice allocation, the system achieves the effect of on-demand allocation of resources, maximizes resource utilization, supports cross-node resource pooling, supports dynamic time-division multiplexing, and sends the task to be processed to the virtual computing node after the allocation of the second time slice so that the virtual computing node after the allocation of the second time slice can be called to execute the task. This solves the problems of inflexible physical deployment, low resource scheduling efficiency, difficulty in integrating heterogeneous computing power, and limited transmission quality caused by electrical signal transmission limitations in traditional computing device resource pooling systems, optimizes transmission performance, and reduces bit error rate and transmission loss. Attached Figure Description

[0017] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a resource scheduling method disclosed in this application;

[0019] Figure 2 This is a flowchart of an optical interconnection transmission method disclosed in this application;

[0020] Figure 3 This is a structural diagram of a resource scheduling system disclosed in this application;

[0021] Figure 4 This application discloses a specific flowchart for implementing resource scheduling;

[0022] Figure 5 This application discloses a schematic diagram of the structure of a resource scheduling device. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0024] In AI large-scale model training and cloud computing scenarios, core problems arise such as low GPU resource utilization, insufficient scheduling efficiency, and physical limitations of traditional electrical interconnect architectures. While existing GPU virtualization technologies can achieve single-machine resource partitioning, the limited transmission distance of PCIe cables makes it difficult to build elastic resource pools across physical devices. Furthermore, electrical signal transmission suffers from drawbacks such as high electromagnetic interference, high bit error rate, and significant power consumption. Therefore, addressing the problems of inflexible physical deployment, low resource scheduling efficiency, difficulty in integrating heterogeneous computing power, and limited transmission quality caused by electrical signal transmission limitations in traditional computing device resource pooling systems, and improving resource utilization efficiency, optimizing transmission performance, and reducing bit error rate and transmission loss, are problems that those skilled in the art need to solve.

[0025] See Figure 1 As shown, this embodiment of the invention discloses a resource scheduling method applied to a resource controller. The resource controller is connected to a target resource pool via optical interconnection, and the target resource pool is configured with several physical computing nodes. Specifically, the method may include:

[0026] Step S11: Obtain the creation request for creating a virtual computing node sent by the target computing platform.

[0027] In this embodiment, a first connection is established between the resource controller and the target computing platform using a cable; based on the first connection, a creation request for creating a virtual computing node sent by the user is obtained from the target computing platform.

[0028] In this application, the target computing platform is connected to the GPU Controller (resource controller) via a cable, and the physical computing nodes are connected to the GPU Controller via optical interconnect (optical fiber and optical module). The target computing platform receives and distributes computing tasks to the physical computing nodes in a software-defined manner. The GPU Controller, as the central device, serves as a bridge for high-speed signal transmission between the general computing platform and the physical computing nodes. It is also used to monitor the target resource pool and schedule GPU resources, forming a unified virtualized resource pool across physical devices.

[0029] Step S12: Select the corresponding physical computing node from the target resource pool according to the creation request, and send the creation request to the physical computing node so that the physical computing node can create the corresponding virtual computing node based on the creation request.

[0030] In this embodiment, an optical fiber and an optical module are used to establish a second connection between the resource controller and the physical computing nodes in the target resource pool via optical interconnection. Based on the creation request, the corresponding physical computing nodes are selected from the target resource pool. The electrical signal corresponding to the creation request is split, and the split electrical signal is converted into an optical signal using the optical module. The optical signal is transmitted to the physical computing node using the optical fiber and the second connection, so that the physical computing node can restore the optical signal to obtain the electrical signal corresponding to the creation request before conversion, so that the physical computing node can create the corresponding virtual computing node based on the creation request.

[0031] In this embodiment, when a creation request for a virtual computing node is received from the target computing platform, the target computing platform sends a vGPU (virtual computing node) creation request to the GPU Controller. After receiving the request, the GPU Controller selects the optimal matching physical computing node that meets the parameters according to the scheduling module, creates the virtual computing node in the target resource pool, and sends a creation success response to the target computing platform.

[0032] The optical interconnect transmission process in this application is as follows: Figure 2As shown, the PCIe signal in the link is converted from an end-to-end electrical signal to an optical signal. This application uses a PCIe switch that supports splitting a PCIe x16 signal into 2x8 signals and a QSFP-DD (Quad Small Form-factor Pluggable Double Density) optical module for photoelectric conversion to form an optical interconnect link. In practical applications, the PCIe x16 signal is split into two x8 signals by the PCIe switch and input into the QSFP-DD optical module. The optical module converts the level information of the PCIe electrical signal into the power information of the optical signal and realizes the optical signal transmission through the optical fiber. Before entering the target resource pool, the optical module at the target resource pool end receives the electrical signal and realizes the optical signal to be restored back to the electrical signal and transmitted to the physical computing node.

[0033] In addition, the resource controller can also use optical switches to establish a third connection with the target resource pool, so as to access the target resource pool based on the third connection, optical switches and optical channels.

[0034] When multiple general-purpose computing hosts are directly connected to an optical switch, each host can access any GPU resource in the target resource pool in parallel through the optical channel. Thanks to the bandwidth advantage of more than 100Gbps per channel and the nanosecond-level latency of fiber optic transmission, the pooled system can still maintain near-local GPU computing performance even in large-scale expansion scenarios.

[0035] Step S13: Forward the creation success response returned by the physical computing node to the target computing platform to obtain the pending tasks sent by the target computing platform.

[0036] In this embodiment, after the physical computing node successfully creates the corresponding virtual computing node, the creation success response returned by the physical computing node is obtained using the second connection relationship; based on the first connection relationship between the resource controller and the target computing platform, the creation success response is forwarded to the target computing platform to obtain the pending tasks sent by the target computing platform.

[0037] This application breaks through the physical boundary limitations of electrical interconnect architecture, enabling ultra-long-distance deployment of GPU nodes; improves resource utilization efficiency, supports cross-node resource pooling, and supports dynamic time-division multiplexing; optimizes transmission performance, significantly reduces bit error rate, reduces transmission loss, and ensures near-local GPU computing performance during large-scale expansion.

[0038] Step S14: Obtain the current status information of the physical computing nodes in the target resource pool, and allocate the corresponding first time slice to each physical computing node based on the current status information.

[0039] In this embodiment, the target resource pool is monitored and managed in real time; wherein, the real-time monitoring includes physical computing node utilization monitoring and network load monitoring; the target time parameters and load status of the physical computing nodes in the target resource pool are obtained in real time, and the corresponding first time slice is allocated to each physical computing node based on the current status information.

[0040] This application allocates a corresponding first time slice to each physical computing node. Physical computing node time slicing is a virtualization technology that allows multiple workloads or VMs (Virtual Machines) to share a single GPU by dividing processing time into discrete slices. Time slices are dynamically allocated based on the usage of each GPU time slice, allocating a portion of the GPU's computing and memory resources to different tasks or users. A time slice reallocation module achieves on-demand resource allocation. This enables the concurrent execution of multiple tasks on a single GPU, maximizing resource utilization.

[0041] Step S15: Based on the required time slice and current status information of the task to be processed, allocate a second time slice to the virtual computing nodes in each physical computing node after allocating the first time slice.

[0042] In this embodiment, a load balancing algorithm is used, and based on the required time slice, target time parameters, and load status of the task to be processed, a second time slice is allocated to the virtual computing nodes in each physical computing node after the first time slice is allocated.

[0043] Specifically, a target ratio is determined; the target ratio is the ratio between the required time slice of the task to be processed and the allocated time slice of any virtual computing node in the target time parameters; it is determined whether the target ratio is greater than a preset threshold; if the target ratio is greater than the preset threshold, a second time slice for any virtual computing node is calculated based on the required time slice and the allocated time slice; the second time slice represents the remaining time slice for any virtual computing node; using a load balancing algorithm and based on the load status, a second time slice is allocated to the other virtual computing nodes in each physical computing node after the allocation of the first time slice, excluding any virtual computing node.

[0044] In this embodiment, the GPU load changes as users utilize the GPU differently. To address these load variations, the time-slice allocation mechanism ensures that the time-slice resources allocated to each vGPU are roughly matched to its load, achieving load balancing among vGPUs. The criterion for dynamically allocating time slices to vGPUs is to first calculate the ratio between the required time slice of the task to be processed and the allocated time slice of any virtual computing node in the target time parameters. Then, it is determined whether the ratio is greater than a preset threshold. For example, the preset threshold is set to 9 / 10. This means that if the percentage of time slices used by a vGPU is less than 90% of its allocated time slices, the remaining time slices of that vGPU will be allocated to other virtual computing nodes besides any other virtual computing node, maximizing the utilization of GPU resources.

[0045] Step S16: Send the task to be processed to the virtual computing node after the second time slice is allocated, so that the virtual computing node after the second time slice is allocated can be invoked to execute the task.

[0046] In this embodiment, after the virtual computing node allocated the second time slice executes the task to be processed, the task execution success response sent by the target resource pool is obtained and sent to the target computing platform; when the resource release task sent by the target computing platform is obtained, the resource release task is forwarded to the physical computing node to be released, so that the physical computing node to be released can release resources of the corresponding virtual computing node based on the resource release task.

[0047] In this embodiment, after receiving the successful creation response, the target computing platform calls the virtual computing node resources to execute the pending tasks. After completing the pending tasks, the target computing platform returns the execution results to the target resource pool and releases the resources of the created virtual computing nodes.

[0048] This application proposes a resource scheduling system, the structure of which is as follows: Figure 3As shown, the system consists of a target computing platform, a GPU Controller (resource controller), and a target resource pool. The target computing platform is primarily responsible for initiating requests to create, invoke, and release GPU resources according to computing and application needs. The GPU Controller manages all GPU resources in the resource pool through GPU monitoring and GPU scheduling modules. The GPU monitoring module monitors GPU utilization, network load, and their status, while the GPU scheduling module is configured with a load balancing algorithm to optimally schedule GPU resources. The target resource pool integrates hardware drivers and runtime libraries from various mainstream manufacturers, using different processes to respond to vGPU resource call API (Application Programming Interface) requests from GPU clients, fulfilling their resource call services, and reclaiming and releasing resources after use.

[0049] The specific process for resource scheduling implemented in this application is as follows: Figure 4 As shown, the core lies in using optical interconnect technology to replace traditional electrical signal transmission and optimizing GPU resource virtualization through time-slice reallocation. By connecting the GPU Controller and GPU nodes (physical computing nodes) via optical fiber, and leveraging the long transmission distance and strong resistance to electromagnetic interference of optical signals, the limitations of traditional PCIe cable transmission distances are overcome, enabling flexible deployment of GPU nodes within a range of hundreds of meters. This architecture physically decouples GPU resources from the CPU host, and the time-slice reallocation module achieves on-demand resource allocation, maximizing resource utilization.

[0050] Furthermore, the optical interconnect technology in this application can realize computing power collaboration across data centers and can be applied to hybrid cloud scenarios, intelligent computing centers, etc. The GPU virtual pooling technology solves the problem of fragmented resource utilization and can be applied to large model inference tasks. Its low latency characteristics can be applied to fields such as autonomous driving.

[0051] Furthermore, if the GPU and optical module operate at full power for extended periods, it leads to energy waste. Therefore, this application also establishes a dynamic energy efficiency regulation mechanism through dynamic GPU frequency reduction and a low-power mode for the optical module, thereby avoiding energy waste. For example, dynamic GPU frequency reduction: when the GPU utilization is below 20%, the core frequency is automatically reduced, such as from 1.8GHz to 1.2GHz, resulting in a 40% reduction in power consumption; low-power mode for the optical module: when the optical link idle time exceeds 5 minutes, the optical module automatically enters "sleep mode," reducing power consumption from 15W to 3W. When data transmission occurs, it wakes up within 100 microseconds, without affecting low-latency requirements. Additionally, a visual panel can be provided, offering users a vGPU resource monitoring page that displays the utilization rate, memory usage, and task progress of allocated vGPUs in real time. It also supports fault alarm push notifications, such as via SMS, email, and WeChat, allowing users to monitor resource status in real time.

[0052] In this embodiment, a creation request for creating a virtual computing node is obtained from the target computing platform; a corresponding physical computing node is selected from the target resource pool based on the creation request, and the creation request is sent to the physical computing node so that the physical computing node can create the corresponding virtual computing node based on the creation request; the creation success response returned by the physical computing node is forwarded to the target computing platform to obtain the pending task sent by the target computing platform; the current status information of the physical computing nodes in the target resource pool is obtained, and a corresponding first time slice is allocated to each physical computing node based on the current status information; based on the required time slice of the pending task and the current status information, a second time slice is allocated to the virtual computing nodes in each physical computing node after the allocation of the first time slice; the pending task is sent to the virtual computing node after the allocation of the second time slice so that the virtual computing node after the allocation of the second time slice can be invoked to execute the pending task. This application applies to a resource controller, which is connected to a target resource pool via optical interconnect. The target resource pool contains several physical computing nodes. The controller receives creation requests for virtual computing nodes from a target computing platform, selects the corresponding physical computing nodes from the target resource pool based on the creation requests, and sends the creation requests to the physical computing nodes. These physical computing nodes are connected to the target resource pool via optical interconnect. Leveraging the long transmission distance and strong electromagnetic interference resistance of optical signals, this application overcomes the transmission distance limitations of traditional high-speed serial computer expansion bus standard cables, enabling flexible deployment of physical computing nodes within a range of hundreds of meters. The physical computing nodes create the corresponding virtual computing nodes based on the creation requests and forward the creation success response returned by the physical computing nodes to the target computing platform to obtain pending tasks sent by the target computing platform. This achieves physical decoupling, significantly reduces bit error rate and power loss, and ensures near-perfect performance during large-scale expansion. The system optimizes the performance of physical computing nodes, obtains the current status information of physical computing nodes in the target resource pool, allocates corresponding first time slices to each physical computing node based on the current status information, and allocates second time slices to virtual computing nodes in each physical computing node after the allocation of the first time slice based on the required time slice of the task to be processed and the current status information. Through time slice allocation, the system achieves the effect of on-demand allocation of resources, maximizes resource utilization, supports cross-node resource pooling, supports dynamic time-division multiplexing, and sends the task to be processed to the virtual computing node after the allocation of the second time slice so that the virtual computing node after the allocation of the second time slice can be called to execute the task. This solves the problems of inflexible physical deployment, low resource scheduling efficiency, difficulty in integrating heterogeneous computing power, and limited transmission quality caused by electrical signal transmission limitations in traditional computing device resource pooling systems, optimizes transmission performance, and reduces bit error rate and transmission loss.

[0053] See Figure 5As shown in the figure, an embodiment of the present invention discloses a resource scheduling device applied to a resource controller. The resource controller is connected to a target resource pool via optical interconnection, and the target resource pool is configured with a plurality of physical computing nodes. Specifically, the device may include:

[0054] The creation request acquisition module 11 is used to acquire the creation request for creating a virtual computing node sent by the target computing platform;

[0055] The creation request forwarding module 12 is used to filter out the corresponding physical computing nodes from the target resource pool according to the creation request, and send the creation request to the physical computing nodes so that the physical computing nodes can create the corresponding virtual computing nodes based on the creation request;

[0056] The pending task acquisition module 13 is used to forward the creation success response returned by the physical computing node to the target computing platform in order to obtain the pending tasks sent by the target computing platform.

[0057] The first time slice allocation module 14 is used to obtain the current status information of the physical computing nodes in the target resource pool and allocate the corresponding first time slice to each physical computing node based on the current status information.

[0058] The second time slice allocation module 15 is used to allocate a second time slice to the virtual computing nodes in each physical computing node after allocating the first time slice, based on the required time slice and current status information of the task to be processed.

[0059] The pending task sending module 16 is used to send pending tasks to the virtual computing nodes after the allocation of the second time slice, so that the virtual computing nodes after the allocation of the second time slice can be invoked to execute the pending tasks.

[0060] In some specific embodiments, the creation request acquisition module 11 may specifically include:

[0061] The first connection establishment module is used to establish a first connection between the resource controller and the target computing platform using a cable;

[0062] The module for obtaining the specific request is used to obtain the creation request for creating a virtual computing node sent by the user from the target computing platform based on the first connection relationship.

[0063] In some specific embodiments, the request forwarding module 12 may include:

[0064] The second connection relationship establishment module is used to establish a second connection relationship between the resource controller and the physical computing nodes in the target resource pool by using optical fiber and optical module and optical interconnection.

[0065] The physical computing node filtering module is used to filter the corresponding physical computing nodes from the target resource pool based on the creation request;

[0066] The electrical signal splitting and conversion module is used to split the electrical signal corresponding to the creation request and use the optical module to convert the split electrical signal into an optical signal.

[0067] The restoration module is used to transmit optical signals to physical computing nodes using optical fibers and a second connection, so that the physical computing nodes can restore the optical signals to obtain the electrical signals corresponding to the creation request before conversion.

[0068] In some specific embodiments, the task acquisition module 13 may specifically include:

[0069] The successful creation response acquisition module is used to obtain the successful creation response returned by the physical computing node after the physical computing node successfully creates the corresponding virtual computing node, using the second connection relationship.

[0070] The successful creation response forwarding module is used to forward the successful creation response to the target computing platform based on the first connection relationship between the resource controller and the target computing platform.

[0071] In some specific embodiments, the first time slice allocation module 14 may specifically include:

[0072] The real-time monitoring and management module is used to monitor and manage the target resource pool in real time; real-time monitoring includes physical computing node utilization monitoring and network load monitoring.

[0073] The target time parameter and load status acquisition module is used to acquire the target time parameters and load status of physical computing nodes in the target resource pool as monitored in real time.

[0074] In some specific embodiments, the second time slice allocation module 15 may specifically include:

[0075] The second time slice allocation module is used to allocate a second time slice to the virtual computing nodes in each physical computing node after the first time slice is allocated, based on the load balancing algorithm and the required time slice, target time parameters, and load status of the task to be processed.

[0076] In some specific embodiments, the second time slice allocation module 15 may specifically include:

[0077] The target ratio determination module is used to determine the target ratio; the target ratio is the ratio between the required time slice of the task to be processed and the allocated time slice of any virtual computing node in the target time parameters;

[0078] The judgment module is used to determine whether the target ratio is greater than a preset threshold.

[0079] The second time slice calculation module is used to calculate the second time slice of any virtual computing node based on the required time slice and the allocated time slice if the target ratio is greater than a preset threshold; the second time slice is the remaining time slice of any virtual computing node.

[0080] The module for allocating a second time slice to other virtual computing nodes is used to allocate a second time slice to other virtual computing nodes in each physical computing node after the first time slice has been allocated, based on the load balancing algorithm and the load status.

[0081] In some specific embodiments, the resource scheduling device may further include:

[0082] The task execution success response sending module is used to obtain the task execution success response sent by the target resource pool after the virtual computing node after the second time slice is allocated executes the task to be processed, and send the task execution success response to the target computing platform.

[0083] The resource release module is used to forward the resource release task to the physical computing node to be released when it receives the resource release task sent by the target computing platform, so that the physical computing node to be released can release the resources of the corresponding virtual computing node based on the resource release task.

[0084] In some specific embodiments, the resource controller uses an optical switch to establish a third connection with the target resource pool, so as to access the target resource pool based on the third connection, the optical switch, and the optical channel.

[0085] The description of the features in the embodiment corresponding to the resource scheduling device can be found in the relevant description of the embodiment corresponding to the resource scheduling method, and will not be repeated here.

[0086] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the resource scheduling method embodiments described above.

[0087] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described resource scheduling method embodiments at runtime.

[0088] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0089] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0090] The resource scheduling method and electronic device provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only intended to help understand the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A resource scheduling method, characterized in that, The method is applied to a resource controller, which is connected to a target resource pool via optical interconnect, and the target resource pool is configured with several physical computing nodes; wherein, the method includes: Obtain the creation request for creating a virtual computing node sent by the target computing platform; Based on the creation request, the corresponding physical computing nodes are selected from the target resource pool, and the creation request is sent to the physical computing nodes so that the physical computing nodes can create the corresponding virtual computing nodes based on the creation request; The creation success response returned by the physical computing node is forwarded to the target computing platform to obtain the pending tasks sent by the target computing platform; Obtain the current status information of the physical computing nodes in the target resource pool, and allocate a corresponding first time slice to each physical computing node based on the current status information; Based on the required time slice of the task to be processed and the current status information, a second time slice is allocated to the virtual computing nodes in each of the physical computing nodes after the first time slice is allocated; The task to be processed is sent to the virtual computing node after the second time slice is allocated, so that the virtual computing node after the second time slice is allocated can be invoked to execute the task to be processed; Based on the required time slice of the task to be processed and the current status information, a second time slice is allocated to the virtual computing nodes among the physical computing nodes after the allocation of the first time slice, including: determining a target ratio; the target ratio is the ratio between the required time slice of the task to be processed and the allocated time slice of any virtual computing node in the target time parameters; determining whether the target ratio is greater than a preset threshold; if the target ratio is greater than the preset threshold, calculating the second time slice of any virtual computing node based on the required time slice and the allocated time slice; the second time slice is the remaining time slice of any virtual computing node; using a load balancing algorithm and based on the load status of the physical computing nodes in the target resource pool, allocating the second time slice to the other virtual computing nodes among the physical computing nodes after the allocation of the first time slice, excluding any virtual computing node.

2. The resource scheduling method according to claim 1, characterized in that, The step of obtaining the creation request for creating a virtual computing node sent by the target computing platform includes: Establish the initial connection between the resource controller and the target computing platform using a cable; Based on the first connection relationship, obtain the creation request for creating a virtual computing node sent by the user from the target computing platform.

3. The resource scheduling method according to claim 1, characterized in that, The step of selecting the corresponding physical computing node from the target resource pool according to the creation request and sending the creation request to the physical computing node includes: By utilizing optical fibers and optical modules, and through optical interconnection, a second connection relationship is established between the resource controller and the physical computing nodes in the target resource pool; Based on the creation request, the corresponding physical computing nodes are selected from the target resource pool; The electrical signal corresponding to the creation request is split, and the optical module is used to convert the split electrical signal into an optical signal; Using the optical fiber and the second connection relationship, the optical signal is transmitted to the physical computing node so that the physical computing node can restore the optical signal to obtain the electrical signal corresponding to the creation request before conversion.

4. The resource scheduling method according to claim 3, characterized in that, Forwarding the successful creation response returned by the physical computing node to the target computing platform includes: After the physical computing node successfully creates the corresponding virtual computing node, the creation success response returned by the physical computing node is obtained using the second connection relationship; Based on the first connection between the resource controller and the target computing platform, the creation success response is forwarded to the target computing platform.

5. The resource scheduling method according to claim 1, characterized in that, The step of obtaining the current status information of the physical computing nodes in the target resource pool includes: Real-time monitoring and management of the target resource pool; real-time monitoring includes physical computing node utilization monitoring and network load monitoring. Obtain the target time parameters and load status of the physical computing nodes in the target resource pool as monitored in real time.

6. The resource scheduling method according to claim 5, characterized in that, The process of allocating a second time slice to the virtual computing nodes among the physical computing nodes after allocating the first time slice, based on the required time slice of the task to be processed and the current state information, includes: Using a load balancing algorithm, and based on the required time slice of the task to be processed, the target time parameter, and the load status, a second time slice is allocated to the virtual computing nodes in each of the physical computing nodes after the first time slice is allocated.

7. The resource scheduling method according to any one of claims 1 to 6, characterized in that, Also includes: After the virtual computing node that has been allocated the second time slice executes the task to be processed, the task execution success response sent by the target resource pool is obtained and the task execution success response is sent to the target computing platform. When a resource release task is received from the target computing platform, the resource release task is forwarded to the physical computing node to be released, so that the physical computing node to be released can release resources of the corresponding virtual computing node based on the resource release task.

8. The resource scheduling method according to claim 1, characterized in that, The resource controller establishes a third connection with the target resource pool using an optical switch, and accesses the target resource pool based on the third connection, the optical switch, and the optical channel.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the resource scheduling method as described in any one of claims 1 to 8 when executing the computer program.

Images