Home

Welcome to the Alnair wiki!

Below is a list of user documents and design documents that can help you quickly get familiar with Alnair.

Build and Setup

Documents

Below are some overview, talks and design documents that will help you understand key features in Alnair.

• Overview

  • Alnair Overview

• Scope

  • Resource utilization improvement
    • GPU sharing + intelligent workload placement
  • Intelligent scheduling (workload placement)
    • co-scheduling, predictive placement, complementary placement
  • Cross-stack multi-functional profiler
    • user transparency, built-in intelligence
  • In-memory distributed caching
    • content based hashing, intelligent shared caching layer
  • Elastic training framework
    • dynamic resource allocation without training interruption
  • Secure container runtime with RDMA support

• Cluster Setup

  • Installation Steps

• Test Results

  • GPU Sharing