The Design & Implementation of Sprites

Summary

Sprites are Linux virtual machines designed as fast, disposable computers with root access, persistent 100GB durable filesystems, automatic sleeping, and low inactive cost. To make creation feel instantaneous, the platform removes user-facing OCI container images and keeps pools of preloaded standard Fly Machines, so a new Sprite mainly performs startup work rather than pulling and unpacking an image. Its storage stack follows a hacked-up JuiceFS model, splitting immutable data chunks from metadata, keeping metadata durable with Litestream, and using a sparse NVMe volume as a dm-cache-like read-through cache for object storage; checkpoints and restores primarily shuffle metadata. This design supports migration and recovery without depending on local storage, while low-cost sleeping and object-backed disks suit interactive work more than warm, containerized production workloads, and billing applies only to storage blocks actually written.

Takeaways

1. 1

   Sprites use standard containers internally, but prewarmed pools avoid the slow image-pull and unpacking work that makes Fly Machine creation take over a minute.

2. 2

   Durable Sprite state lives in S3-compatible object storage; local NVMe caches immutable chunks, so local disk loss does not determine the stored state and workloads can migrate or recover more easily.

3. 3

   Sprites and Fly Machines make different trade-offs: aggressive sleeping and object-backed disks favor interactive workloads, while production applications commonly need warm workloads delivered as OCI containers.