The Saga Pattern Explained
ACID transactions die at the service boundary: no database locks rows across five services. Sagas replace them — a sequence of local transactions with a compensating action per step, so any failure triggers a walk backward. This video shows how a saga executes, what a compensation does, and the two ways to drive it.
Topics covered:
- Why distributed transactions can't use 2PC at scale
- The saga: a sequence of local transactions, each with its own commit point
- Compensating actions: the step that undoes a completed step's effects
- The failure path: walking the saga backward and what "rollback" means here
- Orchestration: a central coordinator telling each service what to do
- Choreography: services reacting to events, with the saga implicit in the flow
- Compensation design: why a refund differs from a rollback
- The guarantees you still get: eventual consistency, not atomicity
Related articles
Microservices Are a Cost, Not a Benefit
Every microservice multiplies pipelines, on-call surfaces, and coordination edges. Here is the operational cost math, when a monolith beats it, and how the strangler fig splits one.
Monolith First: Why Simple Beats Distributed
Before microservices made sense for anyone, they were a cost center pretending to be scale. A review of when a monolith is the right architecture — and when the call for services is actually a cry for discipline.
The Law of Demeter in Practice
Message chains, the law of diminishing knowledge, and why 'don't talk to strangers' prevents the most brittle code you'll ever write.
More in Software Architecture
The Strangler Fig Pattern
How to replace a legacy system without a rewrite — routing by path, strangling features incrementally, and keeping both systems honest.
DetailsDomain-Driven Design Basics
The core of DDD without the ceremony — ubiquitous language, bounded contexts, aggregates, and where the model boundary actually goes.
DetailsService Mesh Explained
What a service mesh actually does — sidecars, mTLS, traffic routing, and observability — and the network plumbing it takes off your plate.
DetailsFeature Flags in Practice
How feature flags work end to end — evaluation at runtime, kill switches, and canary rollouts — plus the operational discipline they demand.
DetailsCQRS Explained
Command query responsibility segregation — separate read and write models, why they diverge, and what you actually gain from the split.
DetailsAPI Gateway Patterns
What an API gateway actually does — routing, auth, rate limiting, and aggregation — and which patterns matter for your traffic.
DetailsHexagonal Architecture, Visualized
Ports and adapters in one diagram — how hexagonal architecture keeps business logic independent of databases, HTTP, and everything external.
DetailsEvent-Driven Architecture Explained
How event-driven systems work — events, topics, and consumers — and what the broker actually guarantees in a distributed system.
DetailsMicroservices vs. Monolith
The honest engineering comparison — deployment boundaries, data ownership, failure domains, and the real costs of splitting a system.
DetailsDepth, delivered weekly
One technical dispatch a week — articles and episode notes before they go public.
One technical dispatch per week. No noise.