Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Cluster GPU resource scheduling system

A resource scheduling and resource scheduling module technology, applied in resource allocation, program control design, instruments, etc., can solve problems such as poor scalability, limited computing power, and inability to carry computing tasks, so as to improve processing efficiency and reduce communication burden.

Pending Publication Date: 2021-07-06
北京蓝耘科技股份有限公司
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in practical applications, a single GPU is often unable to carry complex computing tasks, so it is necessary to form a GPU cluster with multiple GPUs to complete complex computing tasks. Resources are scheduled to complete complex computing tasks, but this method has disadvantages such as limited computing power and poor scalability

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
  • Cluster GPU resource scheduling system
  • Cluster GPU resource scheduling system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present invention belong to the protection scope of the present invention.

[0031] Such as figure 1 As shown, a cluster GPU resource scheduling system according to an embodiment of the present invention includes a GPU master node and several GPU sub-nodes, and the GPU master node includes:

[0032] A task receiving module, configured to receive a task input by a user;

[0033] A task division module, configured to divide the task received by the task receiving module into several subtasks;

[0034] A resource monitoring module, configured to monitor the idle rates ...

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 cluster GPU resource scheduling system, which comprises a GPU main node and a plurality of GPU sub-nodes. The GPU main node comprises: a task receiving module, which is used for receiving a task input by a user; a task division module, which is used for dividing the task received by the task receiving module into a plurality of sub-tasks; a resource monitoring module, which is used for monitoring the idle rates of all the GPU sub-nodes in real time, taking the GPU sub-nodes with the idle rates meeting requirements as available GPU sub-nodes, and measuring the path lengths and the communication rates between the available GPU sub-nodes and the GPU main node, so that all the available GPU sub-nodes with the path lengths and the communication rates meeting the requirements form a GPU scheduling subset; and a resource scheduling module, which is used for distributing the plurality of sub-tasks to the plurality of available GPU sub-nodes in the GPU scheduling subset, so that the available GPU sub-nodes execute the sub-tasks. The parallel computing capability of GPU resources is fully utilized, and rapid parallel processing of large complex tasks can be completed.

Description

technical field [0001] The present invention relates to the technical field of cloud computing, in particular to a cluster GPU resource scheduling system. Background technique [0002] In recent years, the graphics processing unit (GPU) has achieved sustained high-speed development in hardware architecture, and has evolved into a highly parallel, multi-threaded and multi-processing core processor with powerful computing capabilities. Single-instruction multi-threaded architecture increases programming flexibility. GPUs are designed to solve problems that can be expressed as data-parallel computing, that is, the vast majority of data elements have the same data path, and have extremely high computational density, which can hide memory access latency. With its powerful computing capabilities, GPU parallel technology has launched a strong impact on traditional CPU applications, and it has been widely used in popular research fields such as video transcoding, physical simulatio...

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
CPCG06F9/5038G06F9/5066G06F9/5072G06F2209/5021
Inventor 李健
Owner 北京蓝耘科技股份有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com