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 🧵

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.

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.

@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...

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.

What this means for your infrastructure: - 2GB pod limits → 1GB (same perf) - 2x tenant density for multi-tenant SaaS - Edge deployment becomes viable under 128-512MB limits - 2x WebSocket connections per server

Huge thanks to James Snell, Andy Wingo, Joyee Cheung, Dmitry Bezhetskov, Igalia, and Cloudflare for making this happen. Open source collaboration at its best. Full write-up with benchmarks and limitations of this approach 👇 blog.platformatic.dev/we-cut-nodej...