I'm working on a project similar to Letterboxd or Goodreads, where I've dockerized my services and am currently hosting everything on a single VPS. My setup includes the frontend, backend, PostgreSQL database, and Elasticsearch clusters. I'm curious about the scalability of this configuration; particularly, when can I expect the database to become a bottleneck? I'm using a powerful VPS from Hetzner with 48 vCPUs, 192GB RAM, and 1TB SSD. Can anyone share their experience or insights on how far I might be able to scale with this setup? Would I be looking at a capacity of 100k users, 1 million, or even more? And does it matter if the users are concurrent? Just looking for a rough estimate!
5 Answers
It's super tricky to pin down an exact user count. Performance can vary widely based on how each service loads the server. For example, a service using your VPS setup might have different capabilities compared to another that’s more intense. You need to do some real load tests to pinpoint where limits are. Are you getting CPU spikes or RAM usage issues? Knowing how many requests per second your setup can handle should help you scale for future growth, but remember, doubling resources doesn’t always equate to double the users you can support!
Honestly, it really depends on many factors. Each application responds differently to user loads. You could use tools like Azure Load Testing to simulate lots of users and figure out at what point your system starts to lag. What 'slower' means is up to you—deciding if a 3-second response time is acceptable or if you want to keep it under a second is critical as you scale up.
Only concurrent users matter when it comes to server load. You could have a million registered users, but if they aren't active, your load will be zero. That said, with your current VPS, if it's properly optimized, handling hundreds, if not thousands, of concurrent users should be doable. Just keep in mind, programming language efficiency matters too—some languages handle request loads much better than others!
There's a lot to consider when scaling up. For instance, integrating a CDN like CloudFront can help cache content and theoretically eliminate limits based on traffic. But with services like your database and Elasticsearch on the same server, expect I/O to become a bottleneck faster than you think. It's tough to give an exact number, but a rough estimate could be in the range of 42,000 concurrent users, depending on how much each user consumes. Also, don't forget that having a million users who just visit once won't put any load on your server, so concurrent users are really what counts.
The best way to find out is to run a load test. This will give you a clearer idea of how your services perform under different conditions and how many users your VPS can actually handle before things start breaking down.
That makes a lot of sense, load testing seems to be the way to go!
I get what you're saying! But CDNs confuse me for apps since they usually deal with dynamic content. You can still cache some API responses, but not everything is as simple as static sites.