As more and more applications are deployed to the cloud, cost becomes a first class architectural driver when designing and developing applications. In this post I want to explain how you can use cost as an architecture fitness function.
What are architecture fitness functions?
I first encountered fitness functions in the Building Evolutionary Architectures book by Rebecca Parsons, Neal Ford and Patrick Kua. They describe how close an architecture is to achieving an architectural aim. Fitness functions introduces continuous feedback on the architecture level and inform the development process as it happens, rather than after the fact(for example when doing an architecture review).
They help to ensure that a system's architecture aligns with its intended goals, meets the desired quality attributes, and adheres to best practices.
Some examples of architecture fitness functions I used in the past are:
- Latency Monitoring: Ensure that response times for critical services do not exceed a predefined threshold.
- Dependency Analysis: Check that services do not have excessive dependencies or cyclic dependencies.
- Security Scanning: Scan for known security vulnerabilities and ensure that access controls are correctly implemented.
- Service Availability: Monitor service availability and uptime, ensuring that it meets agreed-upon service level agreements (SLAs).
By using architecture fitness functions, development teams can maintain a well-structured and high-quality architecture throughout the development and. evolution of a software system, ultimately leading to more reliable and efficient systems.
How to monitor cost?
Now that we now what a fitness function is, let us have a look at some ways to track cost as an architecture fitness function. Of course there are a lot of infrastructure cost estimation tools available. In this post I’ll show an example using Azure Cost Estimator(ACE) although I heard good things about Infracost as well. ACE supports Bicep/ARM and Terraform and can be run as an integral part of your CI/CD pipeline.
Of course to get good estimations, ACE needs to know the expected usage. This can be provided through usage patterns.
ACE will then use this pattern to generate usage based consumption for the Azure resources and provides better accuracy for the cost estimations.
You can run it by directly using the available executable or by using Docker.
Remark: In the example above we expect that the azure-cost-estimator.exe is available at the path environment variable.