Ok, its ten minutes past 6.00 p.m. I’m on my way from Plovdiv to Sofia just right after the wonderful #jProfessional Plovdiv 2019. Although there are only 130 km between two cities, an express train makes them them for two hours and a half through beautiful Balkan mountaines. Just enough time to summarize some thoughts on microservices I recently had. I still wonder aren’t this thoughts too stupid, but they still bother me.
Just to mention, that this #jProfessionals Plovdiv was EPIC! It totally worth a separate post.
Everybody writes microservices, no matter the reason, for good or for bad. Most of the talks on conferences are about microservices. Every new project is expected to be with microservices architecture. If there is no buzzword “Microservices” nobody from customers would ever listen to you. Some of the juniors I have on my teams say – “Of course microservices! Is there another way? Monoliths? Ah, they are bad! Why? It’s obvious!”
So, technically, I see everybody trying to make their software with this architecture. Or they say the do microservices, although the most projects I’ve seen usually consist of 2–3 fat services. Those projects look like a monolith artificially chunked into pieces that talk REST to each other. People try to decouple this chunked elements as much as they can, but most of the time it’s really obvious that these parts are still logically very coupled and cant live without each other.
I quite often have the following conversation when I consult some project:
– “Ok, you’ve got 40 request per day. The interface is for internal use. You’ve got a team of four developers. Why are you doing six microservices, you’re routing requests with Istio, you’ve got an angular UI”
– “Pretty cool, ah?”
– “How do you spread responsibility in the team, each of you develop his own service?”
– “No, everybody is responsible for all of the code.”
– “What will happen if one of the services will fail?”
– “We throw exception and render code 500.”
– “How do you deploy it?”
– “We’ve got one ubuntu VM on AWS right now. There is docker installed there.”
– “You will run all of the services there?”
– “Yes, and the database.”
– “Will you ever scale?”
– “No.. may be up to 50 requests.”
– “What about security?”
– “Oh.. It will run in secure environment!”
– “What do you mean by secure secure environment?”
– “It will run in intranet.”
– “Ok, now a general question: what is the problem you are trying to solve with microservices?”
– “What do you mean?”
– “Why would you split the app in 6 independent apps?”
– “They are logically split. What’s wrong?”
– “Ok, 50 request per day, it can’t work if one or more services stop… Isn’t a JSF/Vaadin small app just enough for this?”
– “Man! You are such a retrograde!”
I start thinking to myself, well, may be I’m getting old. I believe many C developers think the same about us Java developers – “Look how much overhead they make!!!”.
But still I wonder, is it really worth making everything as microservices? Will this automatically guard you from all of the architecture problems? Or would it create new problems you didn’t have before?
This architecture IMHO should not be considered as absolutely universal. Quite often this causes additional unnecessary overhead. The same time most of the developers I use to know sincerely believe that this is the only way of doing things. They don’t even consider other approaches.
I believe, that we as developers/architects should first ask ourselves – “What are the problems are we going to solve when applying microservices architecture? Do I really need it?”.
Before starting a new project, I personally do a list of answers on a single question: “Why this app should not be a monolith?”. If I find myself at least three technical answers different from “This is modern!”, “Everybody does this!”, “We can do it with one team!”, I start considering microservices as a potential architecture for the requested app.
And every time I start a new project I remember myself the dialog above. Does my app deserve to be microservice–based?
This doesn’t mean, microservices are bad and monolith is good. This means, we as professionals should make the right choice for the right case. It is very hard to give recipes, but it is always a good idea just to think before applying a certain architecture!
By the way, there are two great talks by Daniel Bryant “Seven deadly sins of microservices” and “Seven (more) deadly sins of microservices” (available in youtube.). They are very useful.
The interesting thing is that my current project really nicely suits to microservices architecture. We have several teams doing their own services. The services can run independently and they follow the main idea that “A microservice should be doing one thing, but in the best way!”. There will be pressure and scalability issues. Microservices architecture will help us solve these problems.
But there I have some thoughts about the way a current microservice should be implemented. This will be the subject for the next post.
And yes, train is a nice place to work. Especially if there is an A/C plug.
Oh, here is Sofia. It has beautiful renewed central station.