Distributed systems that survive production
Most of my work is distributed systems: services that have to agree with each other, stay available and keep a consistent record while real companies depend on them.
01
Events are the backbone
A shipment is a sequence of events: quoted, ordered, collected, delivered, approved, invoiced. Modelling the system around those events keeps every service reading from the same history instead of keeping its own copy.
02
Kubernetes and AWS, run deliberately
Containerised services, automated build, test and release pipelines, and observability from the start. None of it is exotic. The value is in doing the ordinary things consistently.
03
Production is the real test
Designing a system is one challenge. Running it reliably every day is another, and it is the one customers notice. Load is tested before limits change, and incidents are written up.
04
An outage is a design review
When something fails, the useful question is what in the design allowed it. Blame finds a person. A review finds the next failure before it happens.
Work, writing and films on this
Written by Pratyush Shivam, Chief Technology Officer at Konfidio.