Hacker News Reader

Comment by TobbenTM | Hacker News Reader

Fully managed so I can spend my time on actually building features. In my case, AWS is my go to cloud, and even with a couple of thousand users, Lambda for compute, DynamoDB for database and SNS+SQS for eventing is costing me less than 5€ per month. Yes, there are risks with serverless if you get DDOSed or whatever, but it’s a risk I’m fine with, and can mitigate with gateways in front if necessary. And Lambdas are not locking me in to AWS since I’m running “full” ASP.Net apps in them, so hosting them on actual compute platforms is an easy switch.

View on HN

Very similar tech stack here, with all the same motivations. Biggest differences are I'm using RDS for the database and have a slightly different approach for serverless compute. I started out with Lambda, but the cold start times were bugging me so I moved production hosting to App Runner, which is the next best thing (eliminates cold start without breaking the wallet). Kept staging environment on Lambda. Also using the "fat lambda" approach with ASP.NET apps, so no lock-in here. Spending about $35/mo total - RDS is the biggest contributor, then App Runner and other small things.

I also have an SSR app for the homepage which required some dynamic functionality. Basic Node app hosted in a Cloudflare worker essentially for free, hitting the App Runner hosted API for data.

Interesting. Did you migrate to a serverless architecture or was your app serverless from the start?

It was serverless from the start, although there isn't a whole lot of serverless unique code in it, apart from some event handler bootstrapping to handle SQS triggered Lambdas. Everything else is just standard .Net that would look the same no matter where it's hosted. Huge fan of separating infrastructure concerns from the rest of the app to not lock yourself in so much.

Snap, I'm using CDK to set it all up too which makes everything pretty easy to manage.

Reply on news.ycombinator.com