Zhivko Todorov
ALL CASE STUDIES

CASE 45 · BRIO · 2023

CLOUDFRONTCACHEEGRESSCOST

CloudFront cache hit rate, doubled.

A news publisher served 9 PB/month from CloudFront with a 42% hit rate. The cache wasn’t broken — the cache keys were. Querystrings, cookies, and User-Agent variations fragmented the cache so badly that the same article was being cached as 40+ distinct objects.

INDUSTRY

News publisher

DOMAIN

COST

DELIVERED

2023

STACK

CLOUDFRONT·CACHEPOLICY·ORIGIN REQUEST POLICY·LAMBDA@EDGE·S3

RESULTS

What changed, by the numbers.

CACHE HIT RATE

89%

WAS 42%

ORIGIN FETCH RATE

−81%

INC. EGRESS COSTS

EDGE LATENCY

−24%

p50, GLOBAL

BILL IMPACT

−$18K/mo

NET, AFTER LAMBDA@EDGE

HOW IT WENT

The cache key analysis was the entire job. Article URLs included `?utm_source=...` querystrings that the upstream marketing system was injecting differently per channel. The same article got cached 14 times for the same content. Cookies for personalisation similarly fragmented the cache.

We rebuilt cache policies to whitelist only the parameters that actually changed the response (`region`, `paywall_status`). Lambda@Edge normalised remaining variations before the cache lookup — stripping marketing tracking, hashing personalisation cookies into a small set of buckets.

Hit rate went from 42% to 89% in the first week. Origin fetches dropped 81%. Egress costs to S3 fell off a cliff. Bill impact net of the Lambda@Edge cost was $18k/month saved.

READY WHEN YOU ARE

Let's get your AWS bill (and architecture) in order.

The discovery call is free. You walk away with at least one concrete idea — even if we never work together.

Or email directly →