GPU (Graphics Processing Unit) time-sharing sharing method and system based on Kubernetes cluster

A time-sharing and clustering technology, applied in the computer field, can solve problems such as performance interference, difficulty in guaranteeing the maximum amount of resources, cluster or node collapse, etc., to achieve the effect of ensuring security and fairness

Pending Publication Date: 2022-07-08
INST OF INFORMATION ENG CAS
View PDF2 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, this method uses virtual multiple GPUs for sharing, which is very unstable in the entire large production cluster, and it is easy to cause the cluster or node to crash
[0008] Most of the current GPU sharing methods are to schedule GPU resources in space, but there are still more or less problems in time-sharing, for example, each task requires GPU memory, and the application amount of GPU memory resources depends on Estimated, it is difficult to guarantee that the maximum amount of resources requested will not be exceeded during operation
Resource overcommitment or overcommitment often causes performance-hindering issues
More seriously, this may make the application fail or crash

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
  • GPU (Graphics Processing Unit) time-sharing sharing method and system based on Kubernetes cluster
  • GPU (Graphics Processing Unit) time-sharing sharing method and system based on Kubernetes cluster

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0035] In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be further described in detail below through specific embodiments and accompanying drawings.

[0036] The specific implementation of this patent is based on the Kubernetes cloud cluster. First, the cluster architecture is briefly introduced:

[0037] The microservice design concept followed by the Kubernetes system architecture is "everything is a service, and everything runs around the service". It adopts the distributed architecture of the control panel (Master) and the work node (Node). One master node manages multiple slave nodes, such as figure 1 shown.

[0038] Master (control panel) is the cluster control node, which manages and controls the entire cluster. Its key processes are as follows:

[0039] (1) API server (API server): It is the entry for cluster control, and it is also the entry for adding, deleting, changing, and c...

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 relates to a GPU (Graphics Processing Unit) time-sharing sharing method and system based on a Kubernetes cluster. The method comprises the steps that a front-end module is arranged in a working node of a Kubernetes cluster, the front-end module is a GPU equipment library installed in a container, and the GPU equipment library schedules use of a GPU in a task by intercepting all APIs related to memory and calculation in a CUDA library; a back-end module is arranged in a working node of a Kubernetes cluster, the back-end module manages tokens between containers and time quotas of the tokens, and time sharing of the GPU is achieved through the tokens and the time quotas of the tokens. According to the method, a GPU time-sharing rule is provided, a GPU equipment library is added into a mirror image to serve as a front end, and a rear-end module is added into a node for control, so that the security and fairness of task execution among a plurality of containers in the node can be ensured.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method and system for realizing time sharing based on Kubernetes cluster GPU resources. Background technique [0002] In recent years, information technology has changed rapidly, and Internet-related technologies have matured at an unprecedented speed, making our daily life and work more and more dependent on computers. At the same time, the rapid development of these technologies has made today's society's quality of computing services The demand for and quantity is increasing day by day. However, those traditional computing models can not meet the current needs, which will naturally lead to new computing models to replace them. Cloud Computing is one of the representatives. It appears in the public's field of vision as a new Internet-based computing model. It can be said that it is a major change in the field of modern science and technology. Cloud computing is regarded a...

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): G06F9/50G06F9/48G06F9/455
CPCG06F9/5027G06F9/4881G06F9/45558G06F2209/5013G06F2209/508G06F2009/4557
Inventor 刘万涛虎嵩林韩冀中
Owner INST OF INFORMATION ENG CAS
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products