Table of Contents

Lab Description

In this lab, you are required to reproduce Lab 0 within the https://www.twcc.ai/user/vms/list environment. The key difference is that this time, it will be executed within VCS/VM (Virtual Compute Service) instances. Furthermore, you need to establish a mini-cluster consisting of a head node, a CPU node, and a GPU node, and implement the monitoring system learned in [Chapter 2] 虛擬化與監控 to monitor the cluster.

Deadline

2026/04/06 23:59 (GMT+8)

SPEC

Instances

3 Ubuntu 24.04 VMs (TWCC VCS)

• A 8 cores VM for controller (headnode) w/public IP address

• A 8 cores VM for CPU node (compute node)

• A 8 cores / 1 gpu VM for GPU node (heterogeneous compute node)

  • Install NVIDIA GPU Driver
  • Don't install Desktop related packages. Only compute ones!

• Network

  • Your virtual network in TWCC VCS shouldn’t overlap with the WireGuard subnet.
  • Reproduce Lab0 on the head node (WireGuard part, not the Dockerfile).

  <aside> <img src="/icons/report_yellow.svg" alt="/icons/report_yellow.svg" width="40px" />

  WARNING

  Notice that you should stop your original container in Lab 0, or you cannot use the same IP address in WireGuard VPN.

  </aside>

OJ Access

• Add judge user for all nodes

  • The judge’s UID should be the same!

  • Login Shell: /bin/bash

  • Public Key:

    ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAILti+rM2UdvYTUgBKihp0b/9o2W+KvJVmA2ex0/u8pCV judge@sa-2025
    

• Revise /etc/hosts on all nodes

  • <headnode-private-IP> controller

  • <cpu-node-private-IP> cpu01

  • <gpu-node-private IP> gpu01

  • An example of /etc/hosts file

    172.16.0.1 controller
    172.16.0.2 cpu01
    172.16.0.3 gpu01