April 4, 2026 · Tim Fraser, Cloud Operations Lead

Preparing Your AWS Infrastructure for Black Friday Traffic

Black Friday traffic doesn't ramp up gradually. It hits your infrastructure like a wall at midnight — or whenever your marketing team sends that email blast. If your AWS setup isn't ready, your customers find out before you do: pages time out, carts fail to load, and sales go to your competitors.

The difference between sites that survive peak traffic and sites that don't usually comes down to preparation done weeks before the event. Here's a practical checklist.

Raise your auto-scaling limits early

AWS applies default service limits to every account. These limits exist for safety, but they'll throttle you during a traffic surge if you haven't planned ahead.

Check these limits at minimum:

• EC2 instance limits — you may be capped at 20 on-demand instances per type. Request increases for the families you use.
• Lambda concurrency — the default regional limit is 1000 concurrent executions. If your checkout uses Lambda, this can be a hard ceiling.
• RDS connections — a db.t3.medium allows roughly 80 connections. That runs out fast under load.

Submit limit increase requests at least two weeks out. AWS reviews them manually, and some take days to approve.

Pre-warm your load balancers

Application Load Balancers scale automatically, but not instantly. If traffic jumps from 100 to 10,000 requests per second in a minute, the ALB may return 503 errors while it provisions capacity.

Contact AWS Support to pre-warm your ALB before the event. Give them the expected requests per second, average request/response size, and the date of the spike. Do this at least 48 hours in advance.

Test your database under load

The database is the most common failure point during e-commerce traffic spikes. Specifically:

• Connection pooling — are you using a connection pool (like RDS Proxy or PgBouncer), or does every request open a new database connection? Without pooling, you'll exhaust connections under moderate load.
• Read replicas — are your product pages reading from replicas, or are they hitting the primary? Every product listing query that goes to the primary is competing with checkout transactions.
• Query performance — run your top 20 queries against a production-sized dataset under simulated load. Queries that take 50ms at normal traffic can take 5 seconds when there are 500 concurrent connections fighting for locks.

Warm your CDN caches

If your product images and static assets are served through CloudFront, make sure those caches are populated before the event starts. A cold cache on Black Friday means every request goes back to your origin server, multiplying the load.

Write a script that requests your top 500 product pages and their associated images through CloudFront. Run it the night before. The CDN will cache the responses, and your origin barely gets touched during the rush.

Run a load test that actually simulates the event

Generic load tests ("hit the homepage 1000 times") don't catch real problems. Your load test should simulate the actual user journey:

• Browse a category page (hits the product API, loads images)
• View a product detail page (database read, inventory check)
• Add to cart (session write, inventory reservation)
• Checkout (payment API call, order write, stock decrement)

Steps 3 and 4 are where most sites fail because they involve writes, external API calls, and transactions. Test these specifically. Use a tool like k6, Locust, or Artillery to simulate 5-10x your expected peak traffic.

Set up monitoring for the day itself

During the event, you need dashboards showing request latency at p95/p99 (not averages), error rates by endpoint (especially checkout), database connection count and query latency, and auto-scaling activity. Set CloudWatch alarms that page someone if error rates exceed 1% or p99 latency exceeds 2 seconds on checkout endpoints.

Check your readiness with one question

This checklist has a dozen items across multiple AWS services. Checking each one manually takes hours of console clicking and CLI commands.

plainfra can audit your Black Friday readiness in a single conversation. Connect your AWS account (read-only) and ask:

> "Are we ready for a 10x traffic spike? Check auto-scaling limits, database connections, and load balancer config."

plainfra checks your service quotas, inspects your auto-scaling groups, reviews your RDS configuration, and flags anything that would fail under load — with specific numbers and recommendations. It takes about 60 seconds instead of a full afternoon.