Xblog postVisualizing Microservice ArchitectureWritten by: Matt McLarty | March 9, 2017 - 2:00 PM
This is part one of a three-part blog series on visualizing microservices. See part two here.
Complexity in software systems is an unavoidable pain. In conference presentations, blog posts and especially in our Microservice Architecture book, we in the API Academy have discussed techniques for dealing with complexity in the context of microservices. In particular, we’ve cited Fred Brooks’ delineation of complexity into the essential complexity that is inherent in the business problems a software system solves and the accidental complexity resulting from implementing that solution.
As I work more closely with organizations who are adopting “API-first” engineering practices and microservice architectures, I find this delineation increasingly useful. Firstly, by separating the complexity in this way, it allows architects and engineers to focus on minimizing the system’s accidental complexity.
The technological innovations associated with microservices help tremendously here: containers, automation tools, API management platforms. But what I really like about this complexity demarcation is that it helps us face an inconvenient truth: no technology—no elusive silver bullet that Brooks refers to in the title of his paper—is going to reduce a system’s essential complexity.
“Many of the classical problems of developing software products derived from this essential complexity and its nonlinear increase with size.” — Fred Brooks, No Silver Bullet—Essence & Accident in Software Engineering
However, what we can do is at least understand and document a system’s essential complexity. To do so effectively is a tall order for most technologists, since the expression of a system’s essential complexity should include no mention of the underlying technology used in the solution.
The essential elements of a system’s essential complexity definition are the tasks and functions that the system performs, as well as the objects and interactions necessary to perform them. To optimize the cognitive load of such a definition, it is also natural to express this definition as a visual model. Even with visualization, software architects still struggle to keep the elements in their models technology-agnostic and at a consistent level.*
I believe the reason Eric Evans’ Domain-Driven Design (DDD) methodology has risen to prominence is that it offers a means of visually defining the essential complexity of a software system. In particular, context maps showing domains, subdomains, bounded contexts and their interrelationships come as close as any method I’ve seen to illustrating essential complexity in a useful way. UML certainly intended to provide a means for defining technology-agnostic application descriptions but I believe many of its models naturally bias toward its object-oriented roots. In fact, if you go too deep into the DDD toolbox I think you run into the same bias.**
At this point, I use a derivative of DDD context maps when defining a system of interconnected microservices. I will go into detail on this approach in a future blog post.
Why is it important to understand and document the essential complexity of a software system and what does this have to do with microservices? As microservice systems grow and evolve, much of their benefit comes from the cohesion of the services themselves. Phil Calçado highlights this need in his blog post, “Microservices & the First Law of Distributed Objects”. Cohesion is the product of determining service boundaries, a necessary step in modeling the essential complexity of the system.
Interestingly, harmoniously and seemingly in contradiction to the delineation of complexity described by Brooks, the technology agnostic context map of a microservice system is somewhat isomorphic to the topology of its implementation, much more so than that of an integrated set of monolithic applications. This could be a coincidence or it could be an indication that we’re on to something with this microservice architecture approach.
This blog post is the first in a series. Next up will be a deep dive into context mapping for microservices, followed by an examination of how to sketch the microservice details themselves.Click to get the link
Return to Homepage