ISA (Independent Systems Architecture) is a collection of best practices based on experience in particular with microservices and Self-contained Systems and the challenges faced in those projects.
Far too often microservices are adopted but the projects then fail to succeed. This set of best practices ensures that common pitfalls are avoided and the promised benefits of microservices are achieved.
However, ISA principles don’t always apply - see the FAQ for details.
The principles use “must” for aspects that cannot be compromised about and must be present in the system. “Should” is used in principles that have a lot of benefits, but can be left out in some cases. The principles described here talk about a system. Usually the IT landscape of an enterprise contains many systems. Each system might be built with a different set of principles or maybe even a completely different approach.
The system must be divided into modules that provide interfaces. Access to other modules is only possible through these interfaces. Therefore modules must not directly depend on implementation details of other modules, e.g. the internal data representation in a database. The rest of the principles define how modules might be implemented, and how ISA differs from other modularization approaches.
Modules must be implemented as separate processes, containers, or virtual machines to maximize independence and enable a separation between Macro and Micro Architecture.
The choice of integration and communication options must be limited and standardized for the system. The integration might be done with synchronous or asynchronous communication, and/or on the UI level. Communication must use a limited set of protocols like RESTful HTTP or messaging. It might make sense to use just one protocol for each integration option.
Metadata, e.g. for authentication, must be standardized. Otherwise the user would need to log in to each microservices. This might be done using e.g. a token that is transferred with each call / request. Other examples might include a trace id to track a call and its dependent calls through the microservices.
Each module must have its own independent continuous delivery pipeline. Tests are part of the continuous delivery pipeline so the modules must be tested independently.
Operations should be standardized. This includes configuration, deployment, log analysis, tracing, monitoring, and alerting. There might be exceptions from the standard if a module has very specific requirements.
Standards for operations, integration, or communication should be enforced on the interface level. For example, the communication protocol and data structures could be standardized to a specific JSON payload format exchanged using HTTP, but every module should be free to use a different REST library/implementation.
NOTE: We created a dedicated page to express the reasoning behind these principles.
You can download a presentation about ISA. It is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License so you can use it for your purposes.