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

Inter-Pod communication method and distributed computing system

A distributed computing and pod technology, applied in the field of machine learning, can solve problems such as low task execution efficiency, and achieve the effect of reducing task running time, improving communication performance, and improving task processing efficiency

Active Publication Date: 2021-05-25
BEIJING SENSETIME TECH DEV CO LTD
View PDF4 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In view of this, the present disclosure at least provides a method for inter-Pod communication and a distributed computing system to solve the problem of low task execution efficiency mentioned above

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
  • Inter-Pod communication method and distributed computing system
  • Inter-Pod communication method and distributed computing system
  • Inter-Pod communication method and distributed computing system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0016]In order to better understand the disclosure of one or more embodiments, one or more embodiments will be disclosed in one or more embodiments in one or more embodiments of the present disclosure. The technical solution is clearly, integrated, apparent, and the described embodiments are merely the embodiments of the present disclosure, not all of the embodiments. Based on one or more embodiments of the present disclosure, one of ordinary skill in the art is in the scope of the present disclosure without all other embodiments obtained without creative labor.

[0017]The present disclosure provides a method of communicating across POD in a distributed computing system, wherein the distributed computing system may, for example, a kubernes system. The system architecture of the distributed computing system is briefly introduced as described below with the Kubernego system. Such asfigure 1 As shown, the system can include: Master 11 and multiple nodes (NODE),figure 1 Two nodes are show...

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 embodiment of the invention provides an inter-Pod communication method and a distributed computing system; each node in the distributed computing system comprises a plurality of GPUs, and a plurality of Pods corresponding to a target task are used for running on the GPU of the target node; the target task comprises a first Pod and a second Pod; the method comprises the steps that a first Pod obtains task resource information, wherein the task resource information comprises a GPU set correspondingly occupied by a target task to which the first Pod belongs, wherein the GPU set comprises a first GPU where the first Pod is located and a second GPU where the second Pod is located; and the first Pod establishes P2P connection with the second GPU through the first GPU according to the GPU set. According to the embodiment of the invention, the task processing efficiency during cross-Pod communication is improved.

Description

Technical field[0001]The present disclosure relates to machine learning techniques, and more particularly to a method of communication between PODs and a distributed computing system.Background technique[0002]With the application and popularity of deep learning, and the urgent needs of the GPU-based high-performance computing resources, more and more tasks begin to use large-scale cloud computing distributed systems, for example, the cloud computing distribution The system can be kubernetes. In Kubernetes, the basic unit of scheduling is a POD, and one or more PODs can be created according to the depth learning task submitted by the user, and these POD schedules are performed on the node (Node) in Kubernetes. Different PODs may communicate with each other during the execution of the depth learning task.[0003]In the current Kubernes system, there is a certain degree of isolation between the POD. When communication between different PODs, even if they may be on the same node, it will ...

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
IPC IPC(8): G06F9/48G06F9/50G06F9/445G06T1/20H04L29/08
CPCG06F9/4881G06F9/5027G06F9/5066G06F9/44505H04L67/104G06T1/20
Inventor 叶志晟吴保东孙鹏颜深根
Owner BEIJING SENSETIME TECH DEV 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