I'm building a mobile application whose primary users will be in Turkey, and I'm planning the production setup for a NestJS backend with PostgreSQL, Redis, Docker, and Coolify. The API handles authentication, CRUD operations, QR-code lookups, notifications, and media metadata; it doesn't involve gaming or real-time audio or video. I'm deciding between Hetzner Cloud in Germany and a VPS or dedicated server located in Turkey. Network tests show roughly 25–40 ms latency from Turkey to Germany. Would the lower latency of a Turkish provider matter enough to outweigh Hetzner's hardware, reliability, and infrastructure? I'm also interested in practical experience running applications for Turkish users from Germany and how you would approach scaling later.
4 Answers
The best choice may eventually be a hybrid setup. You could keep the main services in Germany and add infrastructure in Turkey later if measurements show that it helps. A few operators have found Turkish servers reliable, but the cost and operational trade-offs should be evaluated against actual demand rather than assumed in advance.
Don't optimize the last few milliseconds before you have users. For larger static files, a CDN with an edge location in Istanbul can reduce perceived latency while the API remains in Germany. Focus first on backups, monitoring, deployment reliability, and database performance; those are likely to become problems before a 20 ms difference does.
Hetzner has been stable in practice and is a reasonable place to start. One sensible approach is to deploy a small VM now, keep the setup portable, and move to a larger VM or dedicated machine only when usage justifies it. Use your own domain and DNS from the beginning so changing providers later is straightforward.
I'd choose Hetzner for this workload. A 25–40 ms round trip is generally insignificant for ordinary authentication and CRUD requests, while dependable hardware and infrastructure tend to matter much more. You can revisit the location after you have real traffic and latency data.
That makes sense. I mainly wanted to plan ahead, but starting small and scaling only when the project needs it seems safer.