AI Data Centre · Level 7 of 8
Training job slowdown
Level 7 of 8 in the AI Data Centre: a free data centre exercise at Advanced level. A large training run is slow. Read the utilisation, temperatures and link counters, find the bad optic or the throttling GPU, and fix it without stopping more than you must.
Free · no download. Your work is saved on this device. Updated 18 Oct 2026.
The brief
- Client
- Northfield Compute
- Site
- Hall 2, Greenmoor campus
- Work order
- WO-DC-107
- Task code
- DC-07
- Difficulty
- Advanced
- Type of work
- Fault-finding
A large training run is slow. Read the utilisation, temperatures and link counters, find the bad optic or the throttling GPU, and fix it without stopping more than you must.
Skills you practise
- GPU utilisation and temperature
- Link error counters
- Spine-leaf fault isolation
- Draining a node for maintenance
What the test checks
The acceptance test marks the skills this level practises.
- GPU utilisation and temperature
- Link error counters
- Spine-leaf fault isolation
- Draining a node for maintenance
In this lab
- Rack and cable GPU servers
- Power budgets and cooling
- GPU jobs and a DCIM view
Levels in the AI Data Centre
Previous: Replace a failed part from a ticketNext: Power event drill
Part of the AI Data Centre
Walk an AI data hall, rack and cable a GPU server, and size its power and cooling.
Feeds these internships: AI and Machine Learning, Data Science and Analytics, AI Agents and Automation, Electrical Design.