Overview
Thomas Ptacek from Fly explains how Sprites work under the hood as “disposable computers” that can be provisioned in seconds. The key innovation is using warm pools of pre-provisioned machines running identical containers, eliminating the typical minute-long wait times for new compute instances.
Key Facts
- Sprites are ‘ball-point disposable computers’ - instant compute provisioning within 1-2 seconds
- Use warm pools of unused machines in multiple regions - eliminates up to 60-second wait times for new Fly Machine provisioning
- All Sprites run from a standard container - enables predictable pre-provisioning since every worker knows exactly what container will be needed
- Custom filesystem splits storage into data chunks (on S3) and metadata (local SQLite) - provides ~300ms checkpointing and restores for persistence
- Only charges for data that differs from base image - pay-per-use storage model rather than full container overhead
- Uses Litestream-replicated SQLite for metadata coordination - ensures durability without depending on local storage
Why It Matters
This represents a breakthrough in serverless computing architecture where instant provisioning meets persistent storage, potentially reshaping how developers think about disposable compute resources.