Available capacity
The wrapper will identify servers with spare compute and memory, considering job requirements and each machine’s operating rules.
AI-SERVERS
Flexible Compute
The wrapper is intended to identify available GPU resources and assign suitable compute jobs to them. The aim is to make better use of existing hardware while customers’ primary workloads are idle.
Planned
Compute where capacity is available
We are planning a control layer over existing GPU servers. It is intended to assign suitable jobs based on available resources while respecting each server’s primary use.
Job distribution
The scheduler will assign suitable jobs to available servers. The focus is on jobs that can wait or support interruption and resumption.
Evaluate the outcome
We will assess completed work, energy consumption, operating costs and revenue. Higher utilisation alone does not guarantee better economics.
Better use of existing hardware
Using spare capacity can reduce the need to purchase more servers for the same amount of work. It can make financial and environmental sense when the benefit outweighs added energy use and overhead. Actual savings will need to be evaluated from real operations.
The project is planned. Public job submission is not yet available.