The Azure Architecture Center describes the microservice architecture style as consisting of a collection of small, autonomous services.
- NServiceBus endpoint (service): Each service is an autonomously deployable and scalable unit with a private data store.
- Message bus: The message bus provides an asynchronous, reliable, and fault-tolerant communication channel which decouples the services.
- Gateway: A gateway is a facade which allows user technologies such as web browsers to decouple from service implementations. Gateways may also provide further operational facilities, but do not contain business logic. Azure offers a range of gateway services.
Finding good service boundaries is one of the biggest challenges with the microservice architectural style. Suboptimal boundaries often lead to a lack of data isolation and excessive inter-service communication. This often leads to high coupling between services which implement a business processes, sometimes referred to as a distributed monolith. To define autonomous services, it is crucial to focus on business boundaries rather than technical boundaries.
In this presentation, Udi Dahan demonstrates the process of finding good service boundaries. He explains the challenges of traditional layered architectures and covers an approach that cuts across all application layers, outlining the natural lines of loose and tight coupling. Finally, Udi shows how these vertical slices collaborate using events, enabling flexible and high performance business processes.
Communication between services is much slower and error-prone due to network limitations compared to communication between modules inside a single process. This can lead to higher latency and increased impact of network outages. Asynchronous communication between services helps to mitigate these risks.
Users often need to see and interact with data aggregated from multiple services. Several technologies and patterns help to do this while keeping the services decoupled, such as ViewModel Composition, microfrontends, and GraphQL.
A major benefit of the microservice architecture style, where each service is hosted independently, with its own private data store, is the ability for a team to choose the most appropriate technologies for a given service without impacting other services or the teams working on them. Endpoints built with the Particular Service Platform are generally hosted separately, and each may use a different data store technology. The Particular Service Platform also supports cross-platform integration with systems or components which do not use NServiceBus.
In Azure, services following the microservice architecture style often use the following technology options:
- Fully managed service offerings like Azure Functions and Azure App Services are popular choices for systems requiring scalable hosting environments with minimal management. These services also offer convenient solutions when building web-facing applications.
- Containerized applications can be hosted in managed container orchestration platforms like Azure Container Apps, Azure Kubernetes Service, and more. They can also be hosted in custom managed Kubernetes clusters running on Azure Virtual Machines. Container-hosted systems may use serverless data stores like Cosmos DB and serverless messaging technologies Azure Service Bus.