Research Topics

Cloud Continuum

홈

Research Topics

Cloud Continuum Parallel Learning & Infra Support

Sub Project 1 - A

  • Prof. Eui-Nam Huh

Cloud continuum infrastructure

  • DPU based software defined network for large scale AI support
  • Real-Time, high-efficiency broker on cloud continuum

Learning parameter partitioning on cloud continuum

  • DPU based parameter sharing
  • Cloud continuum self-healing
  • Atomic level hybrid freezing for learning acceleration on heterogeneous GPU cluster

Massive parallel learning architecture on cloud continuum

  • Continuum overlay neural network construction

Parallel Learning Architecture on CC

CC Self-Healing

Continuum Overlay Neural Network

Cost Effectiveness Computing

Sub Project 1 - B

  • Prof. Euiseong Seo

거Training cluster monitoring and flexible scaling tool

  • Distributed cluster monitoring & management system for resource usage pattern analysis
  • Dynamic scaling control for cluster resource efficiency

Elastic scheduling of large scale training cluster for GPU efficiency

  • Elastic scheduling algorithm to improve cluster energy/performance efficiency
  • Co-location support on large scale model training/inference

Cluster Monitoring & Flexible Scaling Tool

Training-Aware Cluster Management

개인정보처리방침

Close

이메일무단수집거부

닫기