DGMY operations teams now watch the computing center through a live digital twin. The model follows power draw, liquid and air cooling, and the jobs running on the cluster.

Try the change before the hall does

A new training queue, a cooling set-point or a maintenance window can be played through on the twin first. If the model shows a hot spot or a power limit, the plan is adjusted before anyone walks into the hall.

One picture for compute and facilities

Compute staff and facilities staff read the same model. A GPU job that will run for days shows up next to the cooling capacity it will need, instead of in a separate ticket.

The approach is the one in our Digital Twins solution.