Skip to content

DC Suite

DC Suite is a control plane for self-service GPU cloud clusters on bare-metal infrastructure. It turns a pool of physical GPU servers into an on-demand cloud: users create right-sized clusters in a few clicks, run workloads on them, and tear them down — while operators keep a single fleet healthy, secure, and accounted for.

  • For users

    Create GPU clusters, run SLURM jobs, apply software templates, build images, watch live metrics and logs, and track your spend.

    User Guide

  • For operators

    Deploy DC Suite, onboard bare-metal machines, wire up SSO, and run the fleet with confidence.

    Operators Guide

What DC Suite does

  • Self-service clusters. Pick a hardware profile and a node count; DC Suite reserves machines from the pool, provisions them from a golden image, forms a working cluster, and hands it back ready to use.
  • Heterogeneous fleets. A single cluster can span multiple hardware profiles (node groups), and the fleet can hold many GPU and CPU SKUs at once.
  • Software templates. Apply curated or custom stacks (schedulers, frameworks, dashboards) to a running cluster. Admins publish global templates; users keep private ones.
  • Golden images. Build, burn-in, and promote machine images through a BUILT → CANARY → STABLE lifecycle so every cluster boots from a known-good base.
  • Observability built in. Live GPU/scheduler metrics and streamed control-plane logs, per cluster.
  • Cost visibility. Metered node-hours and GPU-hours rolled up per organization, in your billing currency.
  • Secure by default. SSO-only sign-in, scoped roles, per-tenant isolation, and short-lived cluster access credentials.

How it fits together

flowchart LR
    U[User / Operator] -->|Console or API| CP[DC Suite control plane]
    CP -->|reserve + provision| FM[Fleet manager]
    FM -->|power + validate| BM[(Bare-metal GPU pool)]
    CP -->|forms| CL[GPU cluster]
    CL -->|metrics + logs| OBS[Observability]
    CL -->|metered usage| COST[Cost reporting]

The control plane exposes a web console and a REST API. When you request a cluster it asks the fleet manager to reserve and validate machines from the bare-metal pool, provisions them from a golden image, and forms a GPU cluster. Usage flows into observability and cost reporting for the life of the cluster.

New here?

  1. Users: start with Getting Started, then Core Concepts.
  2. Operators: start with Architecture, then Installation.
  3. Everyone: keep the Glossary and FAQ handy.