Originally posted at Sanyam Kapoor’s DevBlog
Microservices have suddenly become the hip thing to do. While it would be imbecile to question the potential of microservices, what I am surprised at why wasn’t it popular enough pre-2014. While microservices solves plenty of problems which otherwise would have bugged a seasoned developer, but to keep things in the larger perspective, a microservices approach is not the best way to get started as a potential high performance developer. The path to microservices is hard, very hard.
The sudden rise to popularity
To put things in perspective, have a look at the following Google Trends on the popularity of the term “microservices”.
Interestingly, if we dig deeper into data here, it turns out that the hubs for new age internet companies are the ones that are most infatuating with this term. Now juxtaposed to the above graph, have a look at the following stats from the Global Startup Ecosystem Ranking 2015:
If you have a look at the top half of the countries trending with this search term versus the bottom half of the Ecosystem Ranking above, you will find a clear connection. The newer startup economies tend to be finally accepting the ways of engineering, which more mature economies have already been doing for quite a while now. This brings us to the question,
Are microservices really new?
While microservices might be the latest trending term across various ecosystems, the process is not new. In fact,
Microservices are the natural evolution of large monolithic architectures.
A software is the contract between the business and the client. Businesses tend to optimize processes at various levels in their timeline. It is because of this same optimization process that innovations in software engineering, some revolutionary and some naturally destined, happen.
Microservices is the product of the same evolutionary process which follows the principle of loose coupling and high cohesion. And why not? After all, large teams do need space to move with high velocity. Microservices provide that space and flexibility to make autonomous decisions in the fast paced competitive environment.
Microservices brings separation of concerns to the engineering trying to build the next product, helps clear the state of mind and keeps a clear vision with fast iterative development in perspective. But if you have made the decision to transistion to the loosely coupled microservices architecture, think before you go ahead, because the road is full of immeasurable roadblocks.
Do you really need microservices?
For all the developers out there who have had the privilege to build system designs from scratch, would agree with me that everything starts from a single large monolithic application. A novice programmer might as well start without any existing frameworks and services, eventually moving to build bloated piece of code.
As more people tend to contribute to the core of your business, it becomes difficult to restrict everything to a single codebase. In the desperate need to find space, we branch out when the single monolith starts going out of our hands. Here are a fraction of the problems that will arise during building polyglot microservices:
Polyglot Micro-Service Deployments
Considering a polyglot environment, everybody chooses their own preferred technology stack, deployments will become a mess.
A Million Interfaces
For all the services that have been created, will require a different set of interfaces to interact with. The design of such interfaces will vary from developer to developer which adds additional overhead for the newbie.
In this wild web of inter-service communication, one has to absolutely cautious about how to maintain them, bring them down, upgrade the services and so forth.
Continuous Integration and Continuous Delivery
With microservices in place, one will need an automated infrastructure to test and reliably deploy the new code pushed into the central repository store. This brings us new build architecture challenges which one would never care about if it were a monolith.
Streams of data will start flowing with different services deployed. One needs to plan for the storage, debugging and usability of those logs as well.
Microservices are a trade-off between maintainability and production velocity.
What should you really do?
Don’t fall for the new shiny things and listen to your business first.
This is the least one can do. Apply the generic principles of software engineering, one view at a time -
Fix your module, then your component and then your architecture.
Let your system evolve, microservices will follow thereafter. Microservices is the booby trap you don’t want to fall into!