Check patentability & draft patents in minutes with Patsnap Eureka AI!

Cloud platform vGPU load balancing scheduling method, medium and device

A technology of load balancing and scheduling method, which is applied in multi-programming devices, program control design, instruments, etc., and can solve problems such as unbalanced processing task time and waste of efficient computing unit performance.

Active Publication Date: 2021-05-11
INSPUR SUZHOU INTELLIGENT TECH CO LTD
View PDF2 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, due to the different characteristics of CPUs and vGPUs, this scheduling strategy wastes the performance of efficient computing units due to the unbalanced processing time of processors on CPU-vGPU heterogeneous computing platforms.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Cloud platform vGPU load balancing scheduling method, medium and device
  • Cloud platform vGPU load balancing scheduling method, medium and device
  • Cloud platform vGPU load balancing scheduling method, medium and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0037] refer to figure 1 As shown, the present invention provides a cloud platform vGPU load balancing scheduling method, including:

[0038] S100, dividing the tasks performed by the cloud server into several subtasks;

[0039] S200. Construct a two-way queue, and store the subtasks in the two-way queue;

[0040] S300, the two-way queue assigns subtasks to the CPU from the head of the queue, and the calculation amount of the assigned subtasks is The two-way queue assigns subtasks to the vGPU from the end of the queue, and the calculation amount of the assigned subtasks is Among them, S is the total calculation amount of the task, K 1 is the first proportional coefficient, in the specific implementation process, the first proportional coefficient K 1 The value is 80%-90%, so that most of the total tasks are statically scheduled tasks, and W is the scheduling coefficient. refer to figure 2 As shown, in the specific implementation process, the scheduling coefficient W i...

Embodiment 2

[0049] The difference between embodiment 2 and embodiment 1 is that, refer to Figure 4 As shown, the remaining (1-K 1 )×S tasks are dynamic scheduling tasks, and the CPU or vGPU executes the dynamic scheduling tasks in a balanced manner after executing their respective allocated amounts until the two-way queue is empty. Another feasible way to balance the execution of the dynamic scheduling tasks is: if the CPU finishes executing the allocated amount earlier than the vGPU, start the dynamic scheduling tasks from the head of the queue The two-way queue allocates the remaining subtasks in the dynamic scheduling task to the idle CPU or vGPU one by one; compared with embodiment 1, in embodiment 2, most of the vGPUs are executed during task scheduling After a subtask in a dynamic scheduling task is dynamically allocated to a vGPU, finally, if the CPU executes a fully partially allocated dynamic scheduling task, the remaining tasks are allocated one by one to the idle CPU or vGPU...

Embodiment 3

[0052] Embodiment 3 differs from Embodiment 1 in that, refer to Figure 5 As shown in , the task executed by the cloud server is divided into several subtasks; and the attribute of the amount of threads required to execute the subtasks is obtained;

[0053] The subtasks are sorted according to the number of threads required, and the subtasks are stored in the two-way queue according to the number of threads from low to high. In this way, during allocation, the CPU is often assigned subtasks with a small number of threads, while the vGPU is often assigned subtasks with a large number of threads. Further optimization is achieved to increase the speed of cloud server processing tasks.

[0054] The present invention provides a cloud platform vGPU load balancing scheduling medium, which stores at least one instruction, and executes the instruction to implement the cloud platform vGPU load balancing scheduling method.

[0055] The present invention also provides a cloud platform vG...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a cloud platform vGPU load balancing scheduling method. The method comprises the steps that: a task executed by a cloud server is divided into a plurality of subtasks; a bidirectional queue is constructed, the sub-tasks are stored in the bidirectional queue, the sub-tasks are allocated to a CPU from the head of the queue by the bidirectional queue, the allocation quantity of the allocated sub-tasks is the quantity of allocating the sub-tasks to a vGPU from the tail of the queue in the bidirectional queue, and the allocation quantity of the allocated sub-tasks is S which is the total allocation quantity of the tasks, K1 is a first proportionality coefficient, and w is a scheduling coefficient; and the remaining (1-K1)*S tasks are dynamic scheduling tasks, and the CPU or the vGPU executes the dynamic scheduling tasks in a balanced manner after executing the allocation quantities of the respective static scheduling tasks until the two-way queue is empty. By executing the method, the allocation quantity of the tasks of the CPU and the vGPU can be relatively balanced, and resources of the CPU and the vGPU are fully utilized.

Description

technical field [0001] The present invention relates to the technical field of cloud server scheduling methods, in particular to a cloud platform vGPU load balancing scheduling method, medium and device. Background technique [0002] The virtual host runs on the host of the cloud platform. The virtual host uses the CPU resources of the host. The GPU of the host can provide vGPU resources for multiple virtual hosts through vGPU technology. For the virtual host, the CPU and vGPU of the host host constitute a CPU-vGPU Handle structure. [0003] Generally speaking, the amount of tasks processed by the cloud server is often relatively large. For the CPU and vGPU involved in task processing, effective task scheduling can give full play to the resources of the CPU and vGPU and improve the task processing capability of the cloud server. In the prior art, in the task scheduling strategy of the traditional CPU-GPU heterogeneous computing platform, according to the characteristics of ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06F9/50
CPCG06F9/505G06F9/5044G06F9/5038G06F2209/5017
Inventor 左强
Owner INSPUR SUZHOU INTELLIGENT TECH CO LTD
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com