MSA(Microservices Architecture)

kc@example.com (kc kim) — Thu, 30 Sep 2021 14:54:00 +0900

What Is MSA?

Microservices Architecture.
It emerged to overcome the limitations of Monolithic Architecture, where every system component is integrated into a single project.
It is a framework composed of small, independently deployable services that each perform a specific function.
MSA divides one system into completely independent services that can each be deployed separately.
Each service exchanges data through RESTful APIs and together forms one large service.
It focuses on a single business capability while allowing different technology stacks such as programming languages and databases.

Services can interact only through APIs.
Each service exposes its end-point as an API.
The practical implementation details are abstracted.
Technical details such as internal logic, architecture, programming language, database, and quality management systems are thoroughly hidden behind the service API.

Each service is modularized, and modules communicate through RPC, Message-driven APIs, and similar mechanisms.
Individual services can be developed quickly and maintained more easily.
Each service can use an appropriate technology stack.
Services can be deployed independently.
- Continuous deployment(CD) can also be lighter than in a monolithic system.
Each service can scale out independently according to its load.
- This offers substantial benefits in terms of memory and CPU usage.
Even if one service fails, the entire service does not fail.
Each service can be built with different languages and frameworks.
Services are easy to extend.

It is relatively more complex than a monolithic architecture.
Because services communicate through RESTful APIs, that communication has a cost.
Since services are separated, testing and transaction handling are difficult.
- When communication failures or server load occur, how to maintain transactions must be decided and implemented.
Because calls between services are continuous, debugging and integration testing are difficult.
Deploying to an actual production environment is not easy.

Spring: Netflix Hystrix
When the failure rate of requests sent from a client to a remote server exceeds a certain threshold, the server is considered problematic, and meaningless requests are no longer sent. Instead, errors are returned quickly(fail fast).