Repository navigation
Conversation
Companion code for the AWS Compute Blog post "Scaling RL rollouts with Firecracker on Amazon EC2 metal." Dense single-tenant Firecracker microVMs on EC2 bare metal for RL rollouts and agentic code execution: a host agent (topology discovery, cgroup/jailer orchestration, warm pool, metrics), an in-microVM guest agent, a density sweep harness and analysis tooling, host image bake scripts, and a CDK app that provisions a single metal reference host for the lab.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Initial public release of the companion code for the AWS Compute Blog post
"Scaling RL rollouts with Firecracker on Amazon EC2 metal."
This repository packs dense single-tenant Firecracker microVMs onto EC2 bare
metal for RL rollouts and agentic code execution, with per-microVM CPU pinning
and NUMA-aware placement, and a sweep harness that measures density.
What's included
host-agent/(Go) — topology discovery, cgroup/jailer orchestration, warmpool, Prometheus metrics; runs as a systemd service on the lab host.
guest-agent/(Rust) — in-microVM PID 1 speaking a small vsock text protocol.lab/(Go) — density sweep harness and measurement tooling (labctl,labsweep).scripts/(Shell) — host bootstrap, golden-image bake, rootfs conversion.cdk/(Python) — VPC, IAM, results bucket, one metal reference host from thegolden AMI.
analysis/(Python),examples/— post-processing and an end-to-end dispatch example.Testing
lab/,host-agent/,examples/dispatch/:go build/go vet/go testall pass;gofmtclean.cdk/:py_compileand env-agnosticcdk synthfor both instance types.analysis/,scripts/:py_compileandbash -nclean.cargo build --release --target x86_64-unknown-linux-musl) has not been run locally — no Rust toolchain on theprep machine. Its first execution is this PR's CI run; expect to iterate on the
workflow if it's red.
Scope
This is reference code to illustrate the blog's architecture and reproduce its
measurements — not production-hardened. Features exercised in the blog's lab but
not material to this post's story (snapshot restore, hugepages, arm64/m8g,
instance-store placement) are intentionally deferred to follow-up contributions.