Great article, has been shared with the teams at work! Do you think there’s a certain scale at which it becomes worth moving your functions to docker containers, eg to reduce cost?
Yup, absolutely. It's about looking at your Total Cost of Ownership (TCO) though, not just the operational cost (what you pay AWS), so that threshold is probably higher than you might think.
Your TCO for a containerised application would depend on what skillsets you have in the team - e.g. if you have to hire a new engineer at $10k/month*** to help you build an equivalent containerised solution and maintain it (well, someone's gotta be on-call for your infra and fix issues when it breaks), and you save yourself $5k/month on AWS cost by switching from Lambda to containers, then you made a -$5k/month lost on the move. And, if your feature velocity suffers as a result (e.g. CI/CD pipeline gets more complicated and it takes the feature team devs 10% longer to implement new features), then there's also productivity and opportunity costs to consider.
*** The total cost of an engineer is usually much higher than their salary - there are finders fees for recruitment agencies, pension contributions, bonuses, cost for the office, company laptop, and so on. Typically, an engineer would cost you around 2x their annual salary.
Great article, has been shared with the teams at work! Do you think there’s a certain scale at which it becomes worth moving your functions to docker containers, eg to reduce cost?
Yup, absolutely. It's about looking at your Total Cost of Ownership (TCO) though, not just the operational cost (what you pay AWS), so that threshold is probably higher than you might think.
Your TCO for a containerised application would depend on what skillsets you have in the team - e.g. if you have to hire a new engineer at $10k/month*** to help you build an equivalent containerised solution and maintain it (well, someone's gotta be on-call for your infra and fix issues when it breaks), and you save yourself $5k/month on AWS cost by switching from Lambda to containers, then you made a -$5k/month lost on the move. And, if your feature velocity suffers as a result (e.g. CI/CD pipeline gets more complicated and it takes the feature team devs 10% longer to implement new features), then there's also productivity and opportunity costs to consider.
*** The total cost of an engineer is usually much higher than their salary - there are finders fees for recruitment agencies, pension contributions, bonuses, cost for the office, company laptop, and so on. Typically, an engineer would cost you around 2x their annual salary.
Awesome explanation thank you - those other aspects of total cost of ownership are so often forgotten / miscalculated for sure.