☁️ ECI
Elice Cloud Infrastructure (ECI) is an IaaS (Infrastructure as a Service) platform designed and operated by Elice. It brings together GPU and NPU instances tuned for AI training, inference, and HPC workloads, a high-speed network fabric for multi-node distributed training, and the operational tooling you need to automate everything around them (API, infrastructure as code, access control, monitoring) — all in one place.
Overall structure
ECI resources are isolated inside a Region → Zone hierarchy. Around each virtual machine, network and storage resources are wired together as follows:
Region
└── Zone
├── Virtual network + firewall rules
│ └── Subnet
│ └── Network interface ──── Virtual machine
│ └── Public IP (optional) ── Internet access
│
├── Block storage ── attached to a VM as a disk
├── Object storage ── API access (S3-compatible)
└── Parallel file system (PFS) ── mounted by multiple VMs simultaneously
- A region is a geographical datacenter unit; a zone is an independent resource-isolation domain inside a region. Every resource belongs to exactly one zone.
- A network interface sits between a VM and a subnet. Public IPs attach to the network interface, not to the VM itself.
- Block storage, object storage, and PFS exist independently of VMs, so your data is preserved even if you delete the VM.
If this is your first time
- Quickstart: hands-on from sign-in to SSH access
- Pricing model: decide between spot, on-demand, or reserved
- Glossary: a fast lookup for unfamiliar terms
Categories
| Section | Contents |
|---|---|
| Quickstart | GPU and spot VM quickstarts, access methods |
| Compute | VM lifecycle, instance types, spot VMs, virtual clusters |
| Storage | Block, object, parallel file system, snapshots |
| Network | Virtual networks, subnets, public IPs, firewall |
| Monitoring | Metrics explorer, alerts |
| Tutorials | GPU ML training, LLM inference, InfiniBand benchmark |
| User management | Users, role-based access control, access tokens, audit log |
| Billing | Pricing model, payment methods, usage lookup |
| Terraform | Terraform overview, quickstart, cheatsheet, spot VM guide |
| FAQ | SSH, GPU, billing, storage and other FAQs |
| Support | Support channels and information to include |
| Glossary | Definitions for frequently used terms |
Frequently visited
- PyTorch CUDA compatibility issues: the first roadblock most people hit before training
- When SSH won't connect: password authentication disabled and other checks
- Operating spot VMs: detecting reclamation and saving checkpoints
- Resizing block storage: growing a disk
- Glossary: when you hit an unfamiliar term
Need help?
- Support: support channels and what to include for a quick response