The Edge Stack
A stack with no obvious home region. Code runs next to the user, the database lives at the edge too, and there are no cold starts to design around. Built for global, latency-sensitive products.
- 01HOBackend FrameworksBackend framework
Hono
Tiny, fast framework built for edge runtimes. Runs on Cloudflare Workers with near-zero cold start and a familiar routing API.
- 02TUDatabasesDatabase
Turso
SQLite at the edge with replicas near your users, so reads resolve locally instead of round-tripping to a single primary region.
- 03BEAuthAuth
Better Auth
Open-source, framework-agnostic auth that runs at the edge without a heavyweight managed service or a central region dependency.
- 04CLHostingHosting and edge
Cloudflare
Workers run code in hundreds of locations with no cold start and no region to pick. The runtime the rest of this stack is built to target.
Why this combo
Every pick here exists to kill latency and the single-region bottleneck. Cloudflare Workers run code next to the user, Hono is built for that runtime rather than squeezed into it, and Turso puts the database itself at the edge so reads do not round-trip to one region. Better Auth keeps authentication in the same edge-native, no-lock-in model instead of reaching back to a central service. The result is an application with no obvious home region, which is exactly the point when your users are everywhere and tail latency is the metric you are judged on.
This stack is organized around one constraint: the slowest request from your furthest user. Each piece is chosen because it runs at the edge natively rather than being adapted to it, which is the difference between an app that feels fast everywhere and one that feels fast only near its origin region.
What this costs to start
Cloudflare Workers, Turso, and Better Auth all have generous free tiers that comfortably cover a launch, and the pricing model rewards the architecture: you pay for requests and storage rather than for always-on servers in multiple regions, which is what a globally distributed app on traditional infrastructure would cost. There are no idle instances to fund in every region because the edge runtime has none.
What to swap when you scale
- Write patterns: edge SQLite is read-optimized; if write volume to a single logical database climbs, shard by tenant or region, which Turso’s multi-database model supports.
- Compute limits: Workers have execution limits suited to short requests; long-running jobs move to a queue or a separate container service rather than staying inline.
- Heavier app logic: Hono composes cleanly, so adding middleware for caching, rate limiting, or auth flows does not mean changing runtimes.
Where it is not the right call
If your workload is write-heavy against a single dataset, or you need long-running processes, heavy transactions, or a large existing Postgres schema, the edge model fights you. This stack rewards read-heavy, globally distributed, latency-sensitive products and penalizes the opposite.