Post

We cut @nodejs memory in half with a one-line Docker image swap. No code changes. No new APIs. Just smaller heaps. Here's what happened when we tested pointer compression on real workloads 🧵

17/2/26, 5:05 pm7 reposts47 likes

Whitney McGruder: Author Coach | Publisher

@whitneymg.bsky.social

18 Feb

That’s an impressive win, simple changes with real impact are always the best kind 🚀✨

Alusus Programming Language

@alusus.bsky.social

27 Feb

Nice, but can it reach the memory use of a low-level language? In the graph is NodeJS 20.12.2. More details: coderlegion.com/11988/benchm...

Matteo Collina

@nodeland.dev

17 Feb

V8 has had pointer compression since 2020 (Chrome uses it!), but Node.js couldn't enable it — because all worker threads shared a single 4GB memory cage. @Cloudflare sponsored @Igalia to fix it in V8 with a new feature called IsolateGroups.

Matteo Collina

@nodeland.dev

17 Feb

The fix: each isolate now gets its own 4GB cage. Your main thread gets 4GB, each worker gets 4GB. The old process-wide limit is gone. The Node.js integration by @jasnell is 62 lines across 8 files.

Matteo Collina

@nodeland.dev

17 Feb

@platformatic built node-caged, our Node.js 25 Docker image with pointer compression enabled: and benchmarked it on AWS EKS with a real Next.js e-commerce app (10K cards, 100K listings, SSR, search, simulated DB delays). hub.docker.com/r/platformat...

Matteo Collina

@nodeland.dev

17 Feb

Results: → ~50% memory reduction → 2-4% average latency overhead → p99 and max latency actually improved Smaller heap = shorter GC pauses. For most teams, this is a no-brainer.