Microservices trade design complexity with operational complexity, to address the shortcomings & patterns of the modern IT sector, at both worker and management levels.
This was an excellent choice for consulting (companies like ThoughtWorks, who were pushing microservices) as it optimized their workflows and bottom line: the answer to every client's question of "what's the architecture?" was "Microservices". The companies pimping m.s. were internally optimized for this architecture - running a consulting shop where the time between "here is our project" to "here is our proposal" is approaching zero is profitable.
And the best part -- just like all those "GPT-processes" now being pushed by companies built around an api -- is that by the time clients encounter operational complexity of running micro-services, the consultants have long left the building.
I would argue Microservices trade 100% design complexity with 100% operational complexity and 20% design complexity. Everytime I saw a microserviced design, I always wondered how would they debug this easily and re-operationalize it. Sometimes it is better to let the entire system die and spawn it back instead of having something running to make it look like it's alive.
> Microservices trade design complexity with operational complexity
Yup!
And it annoys me so much how architects/consultants get away with just proposing 'microservices' without being held responsible for all the increased operational complexity. It's left as something for devops to figure out later.
Microservices trade design complexity with operational complexity, to address the shortcomings & patterns of the modern IT sector, at both worker and management levels.
This was an excellent choice for consulting (companies like ThoughtWorks, who were pushing microservices) as it optimized their workflows and bottom line: the answer to every client's question of "what's the architecture?" was "Microservices". The companies pimping m.s. were internally optimized for this architecture - running a consulting shop where the time between "here is our project" to "here is our proposal" is approaching zero is profitable.
And the best part -- just like all those "GPT-processes" now being pushed by companies built around an api -- is that by the time clients encounter operational complexity of running micro-services, the consultants have long left the building.