How do you handle dynamic computing environments?

It is crucial to understand all the dependencies of an application before choosing the best fit from architecture perspective. With the ever increasing cloud trend customers are finding out after the fact that not all applications are suited for cloud. These could range from high performing applications to storage heavy ones, it is imperative to iron out the details specially when it comes to cost prior to choosing an destination. In an ideal scenario a Hybrid set up may be most suitable.

As part of requirements gathering, it's important to get feedback from who the end customer or users of whatever system you are building. Even if you're building an API which another API will be invoking, there are nuances to expected behaviors, failure scenarios, and business impact on successful or poor implementation.

To understand the requirements we should first understand the context and why there is the need of the requirement. We should understand the non functional aspects as they are many times assumed. Example if the requirements spells that the application is critical then try to get information on what does it mean by critical. Is is high impact business need or it is needs to be available for other application function proper.

Choosing right tool is very important when we are architecting a solution in real world. In one of my architecting need application owners have the containers which they need to host on cloud. To provide a better solution, i came up with the decision tree where they can decide which service to use which do the similar functions. Do they really need AKS to host their containers or they can achieve this via Azure container apps or azure container instances or Azure App services.

It is hyper important to involve the end user as part of the QA/Testing phase. As developers we often focus on functionality but may miss key components or workflows the end users experience which could further optimize the application and make for not only a feature rich deployment but one that is also highly adoptable.

Another dimension to building any new service or application is the overall operations of the service. Most modern DevOps/SRE practices are built into the code and deployment itself, and in the real world, not all failure scenarios or bugs can be automated away, but require a human in the loop to do forensics, root cause analysis, and perform mitigation to bring back the service. All this to say, there is an ongoing operational component to any new piece of software that is written. How and Who will maintain it in the future needs to be part of the current / present thinking.