How to Use WorldLand Cloud

This guide explains how to rent GPU resources on WorldLand Cloud as a customer.

Prerequisites

• Wallet (MetaMask or compatible Web3 wallet)
• WL tokens for payment

Step 1: Connect Wallet

1. Visit the WorldLand Cloud dashboard
2. Click Connect Wallet
3. Sign the authentication message
4. Your session will be created

Wallet Authentication

WorldLand uses EIP-712 signature-based authentication. No password required - your wallet is your identity.

Step 2: Browse Available GPUs

Navigate to Providers to see available GPU resources:

Available Providers:
┌────────────────────────────────────────────────────────────────┐
│ Provider ID    │ GPU Model      │ Available │ Price/Hour      │
├────────────────┼────────────────┼───────────┼─────────────────┤
│ provider-001   │ RTX 4090       │ 2/4       │ 0.50 WL        │
│ provider-002   │ RTX 3090       │ 1/2       │ 0.35 WL        │
│ provider-003   │ Tesla T4       │ 4/4       │ 0.25 WL        │
└────────────────────────────────────────────────────────────────┘

Step 3: Create a GPU Job

Option A: Dashboard UI

1. Click Create New Job
2. Select provider or GPU type
3. Configure resources:
   • GPU count
   • CPU cores
   • Memory
   • Storage
   • Duration
4. Set SSH password
5. Click Create

Option B: API

curl -X POST https://api.worldland.cloud/api/v1/jobs \
  -H "Authorization: Bearer <your-session-token>" \
  -H "Content-Type: application/json" \
  -d '{
    "provider_id": "provider-001",
    "gpu_count": 1,
    "cpu_cores": "4",
    "memory_gb": "16",
    "storage_gb": "50",
    "ssh_password": "your-secure-password",
    "duration_hours": 24
  }'

Response

{
  "job_id": "gpu-0x1234-1706234567",
  "status": "creating",
  "gpu_count": 1,
  "gpu_model": "RTX 4090",
  "cpu_cores": "4",
  "memory_gb": "16Gi",
  "storage_gb": "50Gi",
  "ssh_host": "123.45.67.89",
  "ssh_port": 30001,
  "ssh_user": "root",
  "ssh_password": "your-secure-password",
  "price_per_hour": 0.5,
  "expires_at": "2024-01-26T12:00:00Z",
  "message": "GPU container is being created. Check status in a few seconds."
}

Step 4: Connect via SSH

Once the job status is Running:

ssh root@123.45.67.89 -p 30001
# Enter your SSH password when prompted

Step 5: Use Your GPU Container

Your container comes with:

• NVIDIA CUDA drivers pre-installed
• Root access
• Full GPU access

Verify GPU Access

# Check NVIDIA driver
nvidia-smi

# Example output:
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 535.104.05   Driver Version: 535.104.05   CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce RTX 4090  | 00000000:01:00.0 Off |                  Off |
|  0%   35C    P8    20W / 450W |      0MiB / 24576MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

Install Your Tools

# For PyTorch
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

# For TensorFlow
pip install tensorflow[and-cuda]

# For Hugging Face Transformers
pip install transformers accelerate

Step 6: Monitor Job Status

Check Status

curl https://api.worldland.cloud/api/v1/jobs/gpu-0x1234-1706234567 \
  -H "Authorization: Bearer <your-session-token>"

Job States

Status	Description
creating	Container being provisioned
Pending	Waiting for resources
Running	Ready for SSH access
Failed	Error occurred
Succeeded	Job completed

Step 7: Terminate Job

When done, delete the job to stop billing:

curl -X DELETE https://api.worldland.cloud/api/v1/jobs/gpu-0x1234-1706234567 \
  -H "Authorization: Bearer <your-session-token>"

Resource Configuration Guide

Recommended Configurations

Use Case	GPU	CPU	Memory	Storage
Light Inference	1	2	8Gi	20Gi
Model Training	1	4	16Gi	50Gi
Large Model	1	8	32Gi	100Gi
Multi-GPU	2+	16	64Gi	200Gi

Memory Recommendations by Model Size

Model Size	Minimum GPU Memory	Recommended System Memory
7B params	16GB	16Gi
13B params	24GB	32Gi
30B params	48GB (multi-GPU)	64Gi
70B params	80GB+ (multi-GPU)	128Gi

Troubleshooting

OOMKilled Error

If your container is terminated due to memory:

{
  "status": "Failed",
  "failure_reason": "OOMKilled",
  "failure_message": "Container was killed due to memory limit exceeded",
  "suggestion": {
    "action": "increase_memory",
    "recommended_memory": "32Gi",
    "message": "메모리가 부족하여 컨테이너가 종료되었습니다. 32Gi 이상의 메모리로 새 Job을 생성해주세요."
  }
}

Solution: Create a new job with more memory.

SSH Connection Refused

• Wait for job status to be Running
• Verify the correct IP and port
• Check if firewall is blocking the NodePort

GPU Not Detected

# Check if NVIDIA driver is loaded
lsmod | grep nvidia

# Check CUDA installation
nvcc --version

Best Practices

Cost Optimization

1. Right-size resources - Only request what you need
2. Set appropriate duration - Don't over-provision time
3. Delete jobs promptly - Stop billing when done
4. Use spot-like pricing - Check for lower-cost providers

Data Persistence

Container storage is ephemeral. Always backup important data before terminating a job. Use external storage for persistent data.

Next Steps

• Portal Guide - Dashboard walkthrough
• API Reference - Full API documentation