Sandboxes are the latest hot topic when it comes to AI infrastructure as they are a core component for running agents safely. Interestingly, the space is hyper competitive with dozens of startups (Daytona, Modal, E2B, etc) and large companies (Vercel, Cloudflare, etc) competing with similar product offerings.
This level of competition is unusual in the infrastructure space. The reason is that they are all built on top of a decades-old open-source project called Firecracker that was developed by AWS to power their AWS Lambda product offering.
There aren't many tutorials on Firecracker and it's quite a complicated topic that we will cover step by step. While this is the (ultimate) guide
to Firecracker, it is prone to mistakes and revisions so if you notice anything incorrect just let me know on X at @jverre and I'll fix it !
The guides are going to be released part by part as I build an open-source sandbox orchestration tool. We'll cover:
- Gentle introduction to Firecracker
- Getting started - Running our first VM
- Running a basic VM in Firecracker
- Creating your own VM kernel (coming soon)
- Setting up networking for your VMs (coming soon)
- Optimizing for performance
- Running Firecracker securely (coming soon)
- What is the Firecracker Jailer ?
- Running VMs in production safe isolation
- Setting up monitoring and logging