Skip to content
FIRERUN.

The BaaS & deploy-platform newsroom for builders

Subscribe

Netlify

Netlify Moves Edge Functions Off V8 Isolates, Claims 5x Speedup

Netlify now runs Edge Functions in Firecracker MicroVMs, cutting warm median overhead to 5–6ms from 25–40ms, at the same pricing and with no migration.

Hand-drawn editorial illustration for "Netlify Moves Edge Functions Off V8 Isolates, Claims 5x Speedup"

Netlify has moved Edge Functions off V8 isolates and onto Firecracker MicroVMs, cutting warm-invocation overhead to about 5–6ms at the median from 25–40ms, the company said in a Sept. 29 engineering post. Netlify said the new infrastructure is already serving production traffic at the same pricing, with nothing for customers to change.

The numbers

Netlify said about a billion Edge Functions run on its platform every day. On its figures, a warm invocation (routing to a compute node, entering a MicroVM, running the function and producing response headers) now takes about 5–6ms at p50. Netlify reported p99 invocations 47.4% faster, 99.998% availability and 5x faster edge function log delivery.

Cold starts hit about 1.2% of invocations, Netlify said, and average about 9ms. Those happen when a region has not yet fetched a function’s images.

How it works

Requests used to leave Netlify’s network for a hosted execution service and come back. Now the edge node writes a machine spec for each request and forwards it to a compute node inside Netlify’s own network. Each distinct deploy gets its own MicroVM.

Netlify said MicroVMs start in about 2ms at p99. It snapshots each VM once the JavaScript server is listening, scales idle ones to zero and restores from the snapshot on the next call. Lifecycle management comes from Unikraft, which worked with Netlify on the migration.

The company framed isolation as the security gain: a compromised deploy that escapes its runtime still can’t reach other customers’ code. “V8 isolates, no matter their name, do not provide this level of isolation,” the post said.

What Netlify says comes next

Netlify said a full VM filesystem removes most reasons for the caveats on npm packages, which are in beta. It also said it can now revisit the documented limits of 50ms CPU per request, 512MB of memory and 20MB of compressed code. No dates or new limits were given.

The skeptics on Hacker News

The post drew 227 points and 104 comments on Hacker News by Oct. 2. Commenter nchmy doubted that isolates were as slow as Netlify’s 25–40ms baseline, since Cloudflare Workers also run on V8 isolates and respond far faster. Another commenter, CodesInChaos, flagged that restoring VMs from snapshots can duplicate random-number-generator state, a risk for UUIDs and cryptography. Netlify’s post doesn’t address it.

Our take

The 5x figure compares Netlify’s old hosted service against its new one, not isolates against MicroVMs in general. The gain looks like it comes mostly from keeping the whole request path in-network, and Netlify’s own post doesn’t isolate how much each change contributed. Treat it as a platform improvement, not a verdict on isolates. The snapshot question deserves an answer before anyone puts entropy-sensitive code at the edge.

Stay in the loop

Get new articles in your inbox

We'll only email you when a new article drops. Unsubscribe anytime.