Harper · mid-tier search cache

Collapsing the search tail

Akamai keys its cache on the exact query string, so every phrasing of the same shopper intent becomes its own entry — and most are evicted from a shared LRU before they are ever reused. Harper sits behind it keyed on the reduced query, so the tail lands on the head.

Free shipping with $49 purchase  ·  Fast & free store pickup  ·  Take an extra 15% off with coupon Harper demo — not kohls.com
KOHL’S
Search above, or pick one of the phrasings below.

Drive it

Peak — does search survive the surge?

Hit rate is the wrong unit for peak. During peak the question is whether search stays up, and that is a queueing question about an origin with a fixed number of render slots. Below is a simulation of that queue, run twice over identical traffic — once keyed the way a CDN keys it today, once keyed on the reduced query. Only the key differs.

multiplier on a normal-day search rate
concurrent search renders the origin can hold
measured: Akamai serves these with max-age=900
same budget given to both strategies
Simulating…

Traffic replay — exact key vs reduced key

Loading…

What the cache holds

Nothing cached yet — run a search.