Deepnote + VPS: The Ultimate Cloud Data Science Setup

Table of Contents

1. Why Combine Deepnote with VPS?
2. Integration Approaches
3. VPS Provider Comparison
4. Step-by-Step Implementation
5. Cost Optimization
6. Conclusion

1. Why Combine Deepnote with VPS?

Deepnote provides an excellent collaborative notebook environment, but has limitations:

Key Benefits of Adding VPS

• Compute Power 🚀
  Run resource-intensive ML training without Deepnote's limits
• Persistent Storage 💾
  Maintain large datasets between sessions
• Custom Environments 🛠️
  Install specific versions of libraries/tools
• Backend Services ⚙️
  Host APIs, databases, and scheduled jobs

2. Integration Approaches

1: SSH Remote Execution

import paramiko

ssh = paramiko.SSHClient()
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
ssh.connect('vps_ip', username='ubuntu', key_filename='key.pem')

stdin, stdout, stderr = ssh.exec_command('python train_model.py')
print(stdout.read().decode())

2: Remote Jupyter Server

On VPS:

jupyter lab --ip=0.0.0.0 --port=8888 --no-browser --NotebookApp.token='' &

From Deepnote Terminal:

ssh -N -L 8888:localhost:8888 user@vps_ip

Option 3: Database Connectivity

PostgreSQL Example

import psycopg2

conn = psycopg2.connect(
host="vps_ip",
database="ml_data",
user="dbadmin",
password="securepass123"
)

3. VPS Provider Comparison

🏆 LightNode (Recommended)

• Feature Details
• Global Coverage 40+ datacenters worldwide
• Pricing From $0.03/hour (2vCPU/4GB)
• Network 10Gbps with DDoS protection
• Special Features One-click JupyterLab deployment

Feature	Details
Global Coverage	40+ datacenters worldwide
Pricing	From $0.03/hour (2vCPU/4GB)
Network	10Gbps with DDoS protection
Special Features	One-click JupyterLab deployment

👉Visit LightNode

** Vultr (Alternative) **

• 20+ locations
• Hourly billing
• GPU instances available
• Visit Vultr

Visit Vultr

4. Step-by-Step Implementation

1. Provision VPS Instance

Recommended minimum specs:

• 2 vCPU cores
• 4GB RAM
• 80GB SSD
• Ubuntu 20.04 LTS

2. Configure Security

Generate SSH keys

ssh-keygen -t ed25519 -C "deepnote-integration"

Set up firewall

ufw allow 22/tcp # SSH
ufw allow 8888/tcp # Jupyter
ufw enable

3. Install Required Software

Install Miniconda

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh -b -p $HOME/miniconda

Initialize conda

eval "$($HOME/miniconda/bin/conda shell.bash hook)"
conda init

5. Cost Optimization

Best Practices

1. Auto-shutdown - Stop instances when not in use

2. Spot Instances - Use discounted compute when available

3. Storage Tiering - Move cold data to cheaper object storage

Estimated Costs

Resource	LightNode Price
2vCPU/4GB	$0.03/hour
4vCPU/8GB	$0.06/hour
8vCPU/16GB	$0.12/hour

6. Conclusion

Why This Combo Works

• Deepnote for collaboration and exploration

• VPS for heavy computation and persistence

• LightNode for cost-effective global infrastructure

Getting Started

• Sign up for LightNode

• Launch an instance in your preferred region

• Follow the configuration steps above

• Connect from Deepnote and scale your workflows!

💡 Pro Tip: Use Deepnote's environment variables to securely store VPS credentials‘

FAQ | about Deepnote

What is Deepnote?

Deepnote is a cloud - based collaborative data science platform. It enables users to create, edit, and share data notebooks in a seamless manner. It offers features like AI - powered code completion, support for multiple programming languages, and easy integration with various data sources. It also provides different plans to meet the needs of different users, from individual data analysts to large enterprises.

Is Deepnote free?

Deepnote has both a free version and paid versions. The free version is suitable for aspiring data analysts and scientists, supporting 3 editors and 5 projects. It offers AI-assisted code completion, comes with a basic machine having 5GB of RAM and 2 vCPUs, has a 7-day revision history, and provides 100 monthly calls for Deepnote AI features. The paid versions include Pro (costing $9 per editor per month with an annual payment, which offers a 40% discount), Team (costing $39 per editor per month with an annual payment, which offers a 34% discount), and Enterprise with custom pricing. These are respectively suitable for professional individuals, data teams, and organizations with higher machine and security requirements. The paid versions add more advanced features, computing resources, and enterprise-level characteristics on top of what the free version offers.