OK. I said it. Integration is still around, but in different form from my last blog post. So what does modern integration looks like? Looking at how agile scrum has taken over traditional waterfall development framework, by enable shorter deliver cycles, faster feedback, and having the flexibility to rapidly adapt changes. I believe it’s time for traditional integration to be agile again. By breaking up traditional ESB into distributed microservices.
A little recap of what should be in Agile integration:
- Distributed Integration
- Lightweight, support distributed deployment
- Pattern Based Integration
- Reusable Connectors
- Cloud native solutions
- Lean artifacts, individually deployable
- Container based scaling and high availability
- Well defined, re-usable, and well managed end-points
- Ecosystem leverage
Base on these three principle, I was asked to create a reference architecture for it. After giving some thoughts into it, drawing up numerous boxes and diagram, something clicked when I watched a 3 year old playing. And here, this is my answer…
Yup. That is it! Reference architecture of Agile Integration. Architecture is often needed during the design phase of a software project, it provides a structure and backbone for the software project. Once it’s done, it’s done, any architectural modification will be difficult and can bring disastrous result. We all knew “one thing will never change in software development is Change”. Maybe it’s because regulation requirement, market demands or simply learning more about the business domain. So I started to think, can we build an architecture that is flexible for change, and can be shaped into the needs of a project as needed. So base on that concept, this is my reference architecture for Agile Integration, the modern integration application development that enables flexibility from many angles.
There are the major components in my reference architecture diagram for Agile Integration:
PaaS - Provides a foundation for fast pace software development by allowing developer to self-service, as well as enable automated provision.It will also boost operational efficiencies by making the entire environment DevOps ready.
Security/Identity management(IAM) - Handles basic authentication and authorization for application interfaces and platforms.
Automation - Automates both the processes of building applications, and rolling deployments strategies for upgrades. Together with continuous integration tools, it can achieve continuous delivery of the application softwares.
Logging and Tracing - What was suppose to be easy in monolithic world, has now become one of the challenge in distributed microservice architecture for integration. As it has now become harder to have a unified view of how things are doing. It is important to have a way to see logging as a whole. And has tracing capability for different activities.
Containerized packages - Container technology allows us to build immutable portable lightweight application package, that is language agnostic. The configuration are independent from this builds allowing same package to deploy quickly in multiple environment.
Container management - In large scale system it is essential to have a convenient management platform to manage all the containers running the applications. Taking care of the monitoring, discovery, recovery and failover of the running containers.
Microservices - Large applications/services are broken down into easy maintainable pieces and can be independently develop and deploy in distributed environment. Each of them should be built with failure in mind. Microservices
Load Balance/Service discovery/Network management - Microservice are built for flexibility, according to current system load, the load should be automatically balance to the each running instance. All services should be registered and can be found in the system without complex setup. Ideally a proxy endpoint for external users to hide away the deployment complexity.
Layered architecture - Logically organize microservice into a sets of layers, each has it’s own responsibility to avoid duplication of effort and make it easier to replace later on. Currently there are 4 layers
- Gateway layer - provides simple gateway routing capability such as versioning, dealing with different platform of devices.
- Composite layer - important middle tier that handles composition of multiple microservices. The do more complex routing from processing the content data itself and sometimes handle more complex data aggregation or normalization.
- Base layer - like the name which is most likely represents the basic components of the system. Handles data retrieval or business logic processing.
- Anti-corruption layer - This layer handles interface to legacy application or anything that work against microservice quick and flexible principle. This layer is built in protection wall to your system, by doing the transforming job and translate between two very distinct implementation of the system.
Application domains - This one is more like a pattern then architecture. But since we mention about layers in software I also wanted to touch a little bit on this. Because it’s all about reining the microservice into a more organized way. It’s a good practice to have each application domain defined (using same set of data model). And have other external integration microservice to wire between each domains. One other good thing about microservice is that you have the flexibility to move each small microservice to different domains as which it fits better with.
API management - Managing the APIs by reinforced access policy, collecting statistics around the usage of APIs. Also building a healthy eco-system among the developer and users of APIs. Further to allow direct revenue source from the APIs.
There are certainly some limitation in today’s technology to allow a complete pliable architectural form. But I believe this model can bring flexibility into the system without significant impact to the system when changes occur. I will explain more details in my upcoming posts.