How to optimize infrastructure using AWS Serverless

About Me  

My journey began in the energy sector, and later I transitioned into IT. Currently, my team and I are building public and computational clouds in Kazakhstan, and I want to share my experience in the field of cloud solutions.

What is Serverless  

There are three classic models in infrastructure:

1. IaaS — you buy a machine and maintain it yourself.
2. PaaS — you lease a machine, and the company helps with maintenance.
3. SaaS — you simply call a taxi without worrying about the car.

Example: A vending machine operates only when you place an order. At night, it sits idle — and you don't need to pay for its operation. Similarly with serverless: if the load is not constant, costs decrease significantly.

Practice  

I worked in Saudi Arabia on a ticketing platform for concerts. During the peak season, we paid $12,000 per month for Kubernetes and databases. After switching to AWS Lambda, costs dropped to $6,000 per season and $3,000 during the off-season.

Lambda scales automatically: when a request comes in, the function starts, processes it, and finishes. You pay only for the executed request.

Lambda Limitations

— execution time limited to a maximum of 15 minutes
— not all languages are supported
— cold start — delay during the first launch
— complex monitoring and debugging — tools like AWS CloudWatch are required.

Where to Use  

Serverless is suitable for tasks with irregular loads: reports, HR systems, peak events, backups. For example, the accounting department at Aviasales generates reports several times a day — there's no need to keep a container running 24/7; Lambda handles it in minutes.

Serverless reduces costs, accelerates scaling, and allows startups to manage without large DevOps teams. However, it is not a universal solution: under constant load, using clusters is more cost-effective.