Scaling for Black Friday: A Postmortem
How one e-commerce team handled a 12x traffic spike without touching a single line of code.
Last November, one of our e-commerce customers ran their first Black Friday sale with Nodus. They expected a 3x spike. They got 12x.
Here’s what happened — and what didn’t.
The setup
Standard Shopify storefront with Nodus in front for edge caching and DDoS protection. No custom infrastructure, no Kubernetes, no auto-scaling configuration. Just cache headers and trust.
The spike
At 12:01 AM EST, traffic started climbing. By 8:00 AM, they were at 8x normal. By noon — peak shopping hours — they hit 12x.
The origin server never went above 15% CPU. Why? Because 94% of requests were served from the edge. Product pages, category pages, images, CSS, JavaScript — all cached at 320+ locations worldwide.
What broke
Nothing. On-call slept through the night.
The 6% of requests that reached the origin were API calls — cart updates, checkout, inventory checks. These are dynamic by nature and can’t be cached. But 6% of 12x is still less than 100% of 1x. The math worked.
Lessons
- Cache headers matter more than server size. A 2-server cluster with good cache headers outperforms a 20-server cluster without them.
- Edge caching is the cheapest scaling strategy. Adding origin servers costs linear money. Edge cache hits cost fractions of a cent.
- Black Friday should be boring. If your infrastructure team is sweating on November 24th, something is wrong with the architecture.
