Global Latency on Oxygen: Benchmarking the Edge

Introduction

Shopify Oxygen promises fast, global performance for Hydrogen storefronts. But how fast is it really — and how does it compare to Vercel or Netlify?

This post explores p95/p99 latency benchmarks for Oxygen, highlighting cold starts, warm starts, and regional performance across NA, EU, and APAC.

Why Benchmark Oxygen?

• 📊 Client pitches → prove edge performance with real data.
• 🛠️ Engineering budgets → TTFB targets drive architecture choices.
• 🌍 Global brands → APAC latency can make or break conversions.
• ⚡ Comparisons → merchants often ask about Vercel or Netlify.

👉 Benchmarks = trust + contracts won.

Oxygen Latency Realities

• Cold starts → 100–300ms overhead for worker spin-up.
• Warm starts → sub-100ms responses in-region.
• p95/p99 variance → spikes at global distance.
• Quotas → bundle size and subrequests affect runtime stability.

Benchmark Methodology

1. Deploy identical Hydrogen storefronts.
2. Run tests with k6 and Lighthouse CI.
3. Collect metrics across 3 regions: NA, EU, APAC.
4. Compare against Vercel and Netlify deployments.

Results Snapshot

👉 Oxygen excels in NA/EU, needs better APAC edge coverage.

Case Example: APAC Retailer

• TTFB ~350ms on Oxygen vs ~300ms on Vercel.
• Fixed via:
  • Preconnect headers.
  • Aggressive caching for PLPs.
  • Streaming non-critical PDP data.
• Outcome: effective latency cut to ~180ms perceived.

Guardrails

• ✅ Cache PDPs/PLPs aggressively with SWR.
• ✅ Add <link rel="preconnect"> for critical APIs.
• ✅ Profile p95/p99, not just averages.
• ✅ Communicate Oxygen APAC limits to clients.

Conclusion

Oxygen delivers strong performance, especially in NA/EU, but APAC remains a challenge. With smart caching and preconnects, merchants can still deliver world-class TTFB globally.

Benchmark Oxygen, and you benchmark trust.