Recommended EC2 size for a Phoenix API App?

I’m building out a Phoenix app that functions as a GraphQL API for my mobile and web platforms.

In development I’ve been getting good microsecond performances with a T2.micro instance, but as I move into production and expect more users, I’m wondering if I need to increase it and how much I should scale it to.

Anyone have experience in running Phoenix apps on EC2 and mind sharing their setup?

For reference, I’m expecting about concurrent 10,000 users within the first year.

Concurrent…what? Logged in users? Connections that are mostly idle? Connections that are busy? Connections that are practically a firehose of data?

3 Likes

The t2 micro includes 1GB memory, which is usually fine for APIs. You may hit the CPU credit limit as the number of connections goes up.

Are you using autoscaling and load balancing? That makes adding additional instances or changing the instance type very easy, just update the launch config and roll out new hosts.