Access Methods

Multiple ways to access your rented GPU environment through secure VPN tunnel.

Web terminal

Browser-based access

Direct URL access:

http://10.77.x.2:4200 - Web terminal interface
No additional software required
Works on all modern browsers
Full Linux command line access

Features:

xterm.js terminal emulation
Copy/paste support (Ctrl+Shift+C/V)
Resizable terminal window
File upload/download interface
Multi-tab terminal sessions

Terminal usage

Basic navigation:

# Check GPU status
nvidia-smi

# Monitor resources
htop

# Navigate workspace
cd /workspace
ls -la

File management:

# Upload files via web interface drag-and-drop
# Or use command line tools
wget https://example.com/dataset.zip
unzip dataset.zip

# Download results
# Use web interface download button
# Or prepare files in /workspace/downloads/

SSH connections

Standard SSH access

Connection details:

Host: 10.77.x.2
Port: 22
Username: rental
Authentication: Password or key-based

# Basic connection
ssh [email protected]

# With custom port
ssh [email protected] -p 2222

# Using SSH key
ssh -i ~/.ssh/gpuflow_key [email protected]

SSH key setup

Generate key pair:

# Create new key for GPUFlow
ssh-keygen -t ed25519 -f ~/.ssh/gpuflow_key
chmod 600 ~/.ssh/gpuflow_key

Add public key to rental:

# Copy public key
cat ~/.ssh/gpuflow_key.pub

# Add to authorized_keys in rental environment
echo "ssh-ed25519 AAAA..." >> ~/.ssh/authorized_keys

Advanced SSH features

Port forwarding:

# Forward Jupyter to local port 8888
ssh -L 8888:localhost:8888 [email protected]

# Forward multiple services
ssh -L 8888:localhost:8888 -L 7860:localhost:7860 [email protected]

# Dynamic SOCKS proxy
ssh -D 1080 [email protected]

X11 forwarding:

# Enable GUI applications over SSH
ssh -X [email protected]

# Launch GUI applications
firefox &
code .

File synchronization:

# Rsync for efficient transfers
rsync -avz ./local_data/ [email protected]:/workspace/
rsync -avz [email protected]:/workspace/results/ ./results/

# SCP for simple transfers
scp model.bin [email protected]:/workspace/
scp [email protected]:/workspace/output.txt ./

VNC remote desktop

Desktop environment access

Connection details:

VNC Server: 10.77.x.2:5901
Display: :1
Default password provided in rental credentials

VNC clients:

Windows: TightVNC, RealVNC, UltraVNC
macOS: Built-in Screen Sharing, RealVNC
Linux: Remmina, TigerVNC, Vinagre
Mobile: VNC Viewer (iOS/Android)

Desktop connection

Connect via VNC client:

Enter server address: 10.77.x.2:5901
Input VNC password from rental credentials
Select color depth and quality settings
Connect to desktop environment

Web-based VNC:

http://10.77.x.2:6080 - noVNC web interface
Browser-based desktop access
No client installation required
Optimized for touch devices

Desktop environment features

Pre-installed software:

Ubuntu Desktop or similar Linux environment
Firefox web browser
LibreOffice suite
Text editors (gedit, nano, vim)
Development tools (VSCode, git)
Graphics tools (GIMP, Blender)

GPU acceleration:

Direct GPU access for applications
CUDA support for development
OpenGL hardware acceleration
Video encoding/decoding capabilities

Port forwarding

Local port forwarding

Forward remote services to local machine:

# Jupyter Lab access
ssh -L 8888:localhost:8888 [email protected]
# Access via http://localhost:8888

# TensorBoard
ssh -L 6006:localhost:6006 [email protected]
# Access via http://localhost:6006

# Custom application
ssh -L 3000:localhost:3000 [email protected]
# Access via http://localhost:3000

Reverse port forwarding

Expose local services to rental environment:

# Share local web server with rental
ssh -R 8080:localhost:80 [email protected]
# Rental can access via http://localhost:8080

# Database access
ssh -R 5432:localhost:5432 [email protected]
# Rental can connect to local PostgreSQL

Dynamic port forwarding

SOCKS proxy for flexible routing:

# Create SOCKS proxy
ssh -D 1080 [email protected]

# Configure browser to use proxy
# Firefox: Network Settings > SOCKS Host: localhost:1080
# Chrome: --proxy-server="socks5://localhost:1080"

Web service access

Pre-configured services

Jupyter environments:

http://10.77.x.2:8888 - JupyterLab interface
http://10.77.x.2:8889 - Jupyter Notebook (classic)
Pre-installed kernels: Python, R, Julia
GPU-enabled ML libraries ready

AI/ML interfaces:

http://10.77.x.2:7860 - Stable Diffusion WebUI
http://10.77.x.2:8188 - ComfyUI
http://10.77.x.2:3000 - Custom ML dashboards
http://10.77.x.2:6006 - TensorBoard

Development tools:

http://10.77.x.2:8080 - VS Code Server
http://10.77.x.2:3001 - Code-server alternative
http://10.77.x.2:9000 - File manager
http://10.77.x.2:8025 - Email/SMTP testing

Service management

Check running services:

# List active services
sudo systemctl list-units --type=service --state=active

# Check specific service
sudo systemctl status jupyter-lab

# Service logs
sudo journalctl -u jupyter-lab -f

Restart services:

# Restart Jupyter
sudo systemctl restart jupyter-lab

# Restart all web services
sudo systemctl restart gpuflow-services

Custom service startup:

# Start additional services
cd /workspace
python -m http.server 8000 &

# Background processing
nohup python train_model.py > training.log 2>&1 &

File transfer methods

Browser upload/download

Web interface features:

Drag-and-drop file upload
Directory upload support
Progress indicators
Resume interrupted transfers
Bulk download as ZIP

Limitations:

Individual file size: <2GB
Total upload: <10GB per session
Browser timeout on very large files

Command line transfers

wget/curl downloads:

# Download datasets
wget https://example.com/dataset.tar.gz
curl -O https://example.com/model.bin

# Download with progress bar
wget --progress=bar https://example.com/large_file.zip

SCP transfers:

# Upload single file
scp local_file.txt [email protected]:/workspace/

# Upload directory
scp -r ./data_folder/ [email protected]:/workspace/

# Download results
scp [email protected]:/workspace/results.tar.gz ./

Rsync synchronization:

# Efficient sync with progress
rsync -avz --progress ./dataset/ [email protected]:/workspace/dataset/

# Resume interrupted transfers
rsync -avz --partial --progress ./large_files/ [email protected]:/workspace/

# Sync with exclusions
rsync -avz --exclude='*.tmp' --exclude='__pycache__' ./project/ [email protected]:/workspace/

Cloud storage integration

Direct downloads:

# Google Drive (with gdown)
pip install gdown
gdown https://drive.google.com/uc?id=FILE_ID

# AWS S3
aws s3 cp s3://bucket/file.zip /workspace/
aws s3 sync s3://bucket/dataset/ /workspace/dataset/

# Azure Blob Storage
az storage blob download --container-name mycontainer --name myfile.zip

Upload results:

# Upload to S3
aws s3 cp /workspace/results.tar.gz s3://bucket/

# Upload to Google Drive
# Use web interface or gdrive CLI tool

Performance optimization

Connection tuning

SSH optimization:

# ~/.ssh/config optimization
Host gpuflow-rental
    HostName 10.77.x.2
    User rental
    Compression yes
    ServerAliveInterval 60
    ServerAliveCountMax 3
    TCPKeepAlive yes

VNC optimization:

Reduce color depth for faster rendering
Lower screen resolution if possible
Disable desktop effects and animations
Use local VNC client instead of web browser

Network performance

Bandwidth testing:

# Test upload speed
dd if=/dev/zero bs=1M count=100 | ssh [email protected] 'cat > /dev/null'

# Test download speed
ssh [email protected] 'dd if=/dev/zero bs=1M count=100' | cat > /dev/null

# iperf3 testing
ssh [email protected] 'iperf3 -s' &
iperf3 -c 10.77.x.2

Optimize for latency:

Use geographically close providers
Prefer wired internet over WiFi
Close bandwidth-heavy applications
Use compression for text-based protocols

Resource monitoring

Monitor resource usage:

# Real-time monitoring
htop
nvidia-smi -l 1
iotop

# System information
free -h
df -h
lscpu

Network monitoring:

# Monitor network usage
iftop
nethogs
ss -tuln

Security considerations

Access control

Network isolation:

Services only accessible through VPN tunnel
No direct internet exposure
Provider cannot access your local network
Encrypted connections for all protocols

Authentication:

Unique credentials per rental
SSH keys preferred over passwords
VNC passwords auto-generated
No persistent access after rental expiry

Data protection

Sensitive data handling:

# Encrypt sensitive files
gpg --cipher-algo AES256 --compress-algo 1 --s2k-mode 3 \
    --s2k-digest-algo SHA512 --s2k-count 65536 --symmetric file.txt

# Secure file deletion
shred -vfz -n 3 sensitive_file.txt

Network security:

All connections encrypted (SSH, HTTPS, VNC with encryption)
WireGuard provides additional encryption layer
No cleartext protocols used
Certificate validation for HTTPS

Best practices

During rental:

Change default passwords immediately
Don’t store credentials in files
Log out when not actively using
Monitor for unusual activity

Before expiry:

Download all important data
Clear browser cache and saved passwords
Remove SSH keys from local agent
Verify no background processes remain