Netlify Overhauls Edge Compute Architecture with Firecracker MicroVM Transition

Netlify has rebuilt the underlying architecture of its Edge Functions platform, shifting execution from third-party hosted services to Firecracker MicroVMs running natively within its own edge network. As reported via Hacker News, the infrastructure update delivers median execution speeds that the company states are roughly five times faster while enhancing tenant isolation across the platform's daily workload of about one billion function invocations.

The web development platform serves hundreds of thousands of websites, handling tasks ranging from user personalization to cookie checks and authentication routing for clients such as Sunweb and Loto-Québec. Under the previous architecture, requests arriving at Netlify's edge nodes were routed over the internet to an external execution service before returning. Under the new model, compute nodes embedded directly within Netlify's internal network execute functions locally, avoiding the public internet transit.

Netlify engineered the compute overhaul in collaboration with Unikraft. To maintain low overhead, the platform utilizes Firecracker MicroVMs that create in under one millisecond and start up in approximately two milliseconds at the 99th percentile. Netlify mounts function files as uncompressed EROFS images using memory mapping, ensuring virtual machines only read the specific code segments required for execution rather than loading entire software bundles into memory.

Each incoming request triggers the generation of a machine specification detailing CPU, memory, and connection limits, alongside references to three core images covering the runtime, platform software, and customer code. Netlify hashes this specification with site-specific metadata to establish a distinct service identifier. This structure enforces isolation, ensuring separate code deployments or environment variables never share a virtual machine instance.

Request distribution relies on rendezvous hashing to route identical service identifiers to specific compute nodes, maximizing cache retention and keeping virtual machines warm. To prevent high-volume workloads from creating hardware hotspots, Netlify relaxes node stickiness once traffic surpasses defined thresholds, spreading heavy traffic across a slice of compute nodes without degrading performance for adjacent services.

MicroVMs automatically snapshot state once their internal JavaScript server opens a port, allowing inactive instances to scale down to zero compute resources when idle. Subsequent requests restore execution state from memory-mapped snapshots. During warm invocations, the complete processing pipeline adds roughly six milliseconds. Cold starts occur on approximately 1.2 percent of requests, taking an average of nine milliseconds to pull necessary images to new regional nodes.

Additional operational enhancements include placing local DNS resolvers directly on compute nodes, deploying expanded telemetry to monitor boot sequences and user code startup times, and setting up automated circuit breakers to quickly isolate malfunctioning compute hardware. A dedicated control plane continuously tracks node health, enabling Netlify to execute zero-downtime infrastructure updates by spinning up parallel node fleets before shifting live traffic.

Netlify confirmed that the infrastructure upgrade is live across its production network with no changes to pricing or developer workflows. Existing declarations, local development setups, URL imports, npm dependencies, and Node.js built-ins remain operational without requiring code modifications. The company stated that running the compute layer directly provides architectural flexibility for future platform capabilities.