Studio Pret
We moved a fashion store to Cloudflare edge nodes, keeping the database online under 3 million daily requests.
handled 3 million requests a day without a single crash.
processed 1,000 transactions an hour during peak collection launches.
System Bottlenecks
During product drops, Studio Pret's legacy SQL database froze under concurrent connection loads. Customers experienced hung checkout pages, which led to abandoned carts and lost sales.
Architectural Move
We rebuilt the storefront in Svelte and migrated the API to Cloudflare Workers. We cached the product catalog in Cloudflare KV to remove read loads from the main database, and stored live cart data in D1 SQLite databases near the user.
Execution Path
Locating Bottlenecks: We simulated concurrent checkout runs on the legacy server to identify the database lockup point.
Schema Refactoring: We rewrote the SQL queries to make them compatible with serverless D1 SQLite.
Frontend Build: We replaced the bloated legacy frontend with a lightweight Svelte page to reduce browser rendering time.
Live Cutover: We shifted traffic to the new edge backend without taking the site offline.
"We used to lose sales during every collection launch because the site crashed. Now the server does not lag at all."
— Ayesha Khan, Operations Director at Studio Pret
Core Stack
- Svelte & SvelteKit
- Cloudflare Workers (Edge Compute)
- Cloudflare D1 (Serverless SQLite)
- Drizzle ORM for database migrations
Under-the-hood Challenge
Keeping database transactions safe across edge servers without adding round-trip delay. We solved this by staging changes on the edge node nearest to the user and writing batch updates in single transactions.
Studio Pret sells apparel online. This case study shows how moving database reads to the network edge keeps a website responsive during high-traffic drops without scaling expensive servers.