I’m invesigating the difficulty of migration from Gigalixir Standard tier to Fly.io.
Actual Gigalixir pre-launch setup for 45€/month includes:
stack: gigalixir-24
region: europe-west
replicas: 1
memory size: 0.4GB
Postgres DB size: 0.6 (25€/month,RAM 0.6 GB, 25 connections, storage 25GB). When go live: move to the next size 1.7 (50€/month, RAM 1.7 GB, 55 connections, storage 50GB). Standard tier databases include 7 days of automatic daily backups.
I tried to play a bit with the price calculator available on the fly.io doc pages but still have no clear picture. Can anybody explain what fly.io plan would more or less correspond to Gigalixir setup?
Using Gigalixir’s calculator, I specified 0.5 GB of RAM, 1 replica, no database, which is $25/mo. That would start from $7 in Fly, but there are a lot of CPU options, and a few people with demanding apps anecdotally seem struggle with the time-sliced CPUs. That may also apply to some folks who have apps that have expensive init routines, such as those that run on the JVM (search for “throttling” here to understand how this works).
I’d say also that even if it were possible to compare providers, you have to look at cloud reliability, cost of support, support workload, etc. The best approach, if your project is business-critical, is to get it to run on multiple providers, so that it is not onerous to move if you need to. However, often, this is easier said than done.
Would it be a lot of work for you to set your project up on Fly in non-prod, and then use load-testing tools over a month, to see if it’s worth a real lift-and-shift?