// on serverless
Serverless isn't a deployment strategy — it's a way of thinking. When you stop managing infrastructure and start composing events, everything changes. Systems become simpler, costs become rational, and you spend your time on what matters: the product.
// the mindset shift
Traditional architecture asks 'what URL should I call?' Serverless asks 'what happened, and what should happen next?' This is a fundamental shift. An SQS message, an EventBridge event, an S3 upload — these are triggers, not requests. When you think in events, your systems become naturally decoupled, resilient, and scalable.
A Lambda function that runs twice a day costs almost nothing. That same logic on a running server costs 24/7. Serverless flips the economics of software — you pay for execution, not existence. This changes which ideas are worth building. Experiments become cheap. Side projects become free.
Lambda has a 15-minute timeout. DynamoDB has a 400KB item limit. These aren't limitations — they're guardrails that force you toward better patterns. Small, focused functions. Efficient data models. Event-driven decomposition. The constraints of serverless push you toward the architecture you should have been building all along.
// the building blocks
The compute layer. Small, focused functions that do one thing well. No servers to manage, no capacity to plan. Write the code, deploy it, forget about infrastructure.
The communication layer. Queues for reliable processing, topics for fan-out. Decouple producers from consumers. Handle spikes by buffering, not by scaling servers.
The nervous system. Route events between services based on rules. Connect AWS services, SaaS applications, and your own microservices through a single event bus.
The data layer. Single-digit millisecond reads at any scale. Design your access patterns first, model your data second. It rewards thinking differently about data.
The front door. REST and WebSocket APIs without managing any infrastructure. Authentication, rate limiting, and request transformation built in.
The blueprint. Infrastructure as code, version-controlled and repeatable. Every environment is identical because it's defined in the same template.
// in practice
In the pension sector, I've built serverless systems that process thousands of transactions daily. An event lands in SQS, a Lambda picks it up, processes it, writes to DynamoDB, and publishes the result to EventBridge for downstream consumers. Each step is independent, observable, and retryable. When something fails, it goes to a dead-letter queue — not into the void.
Legacy systems don't speak REST. They speak SOAP, flat files, and batch jobs. Serverless is the perfect middleware layer — Lambda functions that translate between old and new, triggered by schedules, events, or API calls. You don't need to replace the legacy system. You need to give it a modern interface.
When you move from always-on servers to event-driven functions, your cloud bill starts to reflect actual usage. Low-traffic services cost pennies. High-traffic services scale automatically. No capacity planning meetings, no over-provisioning 'just in case.' The infrastructure matches the workload, minute by minute.
Want to talk about serverless architecture or event-driven systems?