Machines

Machines are allocated to your organization by Acasia — you can’t create or remove them yourself. A machine may represent:

• A single GPU server
• A DGX-class multi-GPU system
• A rented GPU instance
• A partner-supplied host
• Another GPU-backed runtime environment

​

Machine fields

​

Machine vs. cluster

A machine is the underlying hardware. A cluster is a deployable allocation carved from that hardware. One machine can host multiple clusters, each with its own slice of GPU, CPU, RAM, and storage. You create and manage clusters — not machines. Machines are managed by Acasia and appear in your portal once allocated.

​

Clusterization

Some machines support clusterization — the ability to split a single machine’s resources into multiple independent clusters. This allows different teams, workloads, or projects to use isolated compute environments on the same physical hardware. When a machine supports clusterization:

• Multiple clusters can be created from the same machine
• Each cluster has its own resource allocation (GPU, CPU, RAM, storage)
• Clusters operate independently

When a machine does not support clusterization, it maps one-to-one with a single cluster.

​

Machine status

​

Capacity planning

Key questions:

• How many GPUs does the machine have?
• How many clusters already exist on this machine?
• Does the remaining capacity support the planned workload?

If your organization has exhausted available machine capacity, contact Acasia to request additional allocation.