For comparation, 2 years ago we build thumbor-like apps in python using vibora and opencv. A single aws m8 large($80/month?) can serve hundred millions images

For comparation, 2 years ago we build thumbor-like apps in python using vibora and opencv. A single aws m8 large($80/month?) can serve hundred millions images resizing in a month. At that time we build it with only 2 people, copy pasting code from internet in a few days and vibora’s github readme also state that its in alpha release. With $2K server, you need to cheat to make those inefficient implementation :) Side note: Several weeks later I found out about thumbor and replace our ‘emergency work’ with dockerized thumbor I also tinkered with thumbot in aws lambda. But the cost raised to $500/month. I think we did something wrong with aws rekognition.