Skip to main content

☁️ Elice Cloud

Elice offers two cloud services for AI and HPC workloads. They differ in isolation model, level of control, and operational overhead — pick the one that fits your workload.

Service comparison

ItemECI (Elice Cloud Infrastructure)Elice AI Cloud
Isolation unitVirtual machineContainer
OS / runtime controlFull sudo access — install kernel modules, Docker, the NVIDIA container runtime, and other components freely inside the VMPre-configured by Elice — standard runtimes (Jupyter, VSCode, etc.) ready to use
NetworkingBuild your own virtual networks, subnets, public IPs, firewalls, and InfiniBand / RoCEv2 fabricsManaged by the platform. You only expose external ports (HTTP / TCP) to containers
StorageBlock storage (with snapshots and schedulers), object storage, parallel file system (PFS)Block storage, Data Hub (S3-compatible object storage)
Automation interfacesTerraform provider (elice-dev/eci), CLI (eci), REST APIPortal UI
Pricing modelsOn-demand, reserved, spotOn-demand, reserved
Add-on servicesVirtual clusters (HPC), monitoring & alerts, RBAC, activity logsML API (serve trained models as REST endpoints)

How to choose

Pick ECI when you need to

  • Control custom OS, driver versions, or kernel modules directly
  • Run your own container runtime (Docker, containerd, etc.)
  • Train distributed workloads across multiple nodes over InfiniBand (e.g., large-scale LLM pre-training)
  • Automate with Terraform / CLI as IaC, or operate multiple environments (dev / stage / prod)
  • Design network topology precisely — virtual networks, firewalls, VPNs, and so on

Pick Elice AI Cloud when you want to

  • Start training or experimentation immediately in a pre-configured environment
  • Use standard tools like Jupyter or VSCode without managing container builds or runtimes
  • Serve a trained model quickly through ML API
  • Minimize operational overhead

The two services are separate products and do not share resources. A single organization can use both side by side.


Next steps