π Monolithic and Its issues
π When we start learning backend development and implementation, many of us implement all our functionalities in one codebase. Forex. If we want to build the backend of an e-commerce store then we can go with the project with a single index.js file including the functionality of:-
- Shopping Orders
- Items or Products
- Payment
- Store
- Wallet
- Cart
π But is it right to do so? We do have the advantages of the monolithic architecture like easy testing, and easily deployable. But we do have some issues like scalability, load balancing, and error handling.
π― With Microservices we can address these issues and we also can use some of the advanced technologies like Docker, Kubernetes effectively.
π Microservices
π― Microservices architecture includes working of separate services that are loosely coupled and work on different ports and with their own database generally.
π Let's take an example of an eCommerce app with payment, product, and shipping service. It may look like this:-
Example:-
Product
--> Used:- JAVA, MYSQL
--> Working on:- PORT 3001
Payment
--> Used:- NODE, MongoDB
--> Working on:- PORT 4000
Shipping
--> Used:- Python, SQLite
--> Working on:- PORT 3002
π Now these services can communicate with each other using APIs and all. We have the advantages of load balancing, error handling (If payment service fails it will not affect the product service), and modular code with this architecture (Microservices).
π Microservices Patterns
π₯ While we implement microservices, there are some patterns like using BFF (Backend for frontend), Service Discovery
BFF (Backend for frontend)
π Here, we have API Gateway that handles all the stuff like Authentication, load balancing ,etc and act as a reverse proxy to accept the incoming requests and gives the appropriate results.
Service Discovery Registry
π It is used when we have multiple instances of a service and we get the address of the particular instance from the service discovery.
π Note:- We will see the implementation too, Don't get panic if you are not able to understand fully.
Inter-Microservice Communication
π This is one of the method where we can make calls between the services. You can make Sync calls or Async calls depending on the functionality. Here, we have used the Queue where the messages will be passed from service 1.
Circuit Breaking
π This situation arises when one or more calls fails in between and other keeps waiting. forex. Service2-->Service3 call fails and Service 1 will keep waiting. In that case, we can do following:-
- Use the last recent call stored inside the cache from service 2 and 3 and give the response
- Fallback call:- It is the case when Service 3 calls another third party API and waiting for a long time. Then Service 2 can directly call to that 3rd party API
Log Aggregation Pattern
π It is the Pattern where we store all the logs of our active microservices so that we can track and monitor them
Serverless
π It is a cloud computing application where we can use Functions that triggers while we call. These functions do have a cold start time though but might be very effective in most of the applications.
π₯ Follow for more. We will see the implementation in our next Blog. Share this series and drop your views
Oldest comments (14)
Good π
Impressive
Very Helpful. Thanks :)
Thanks for sharing.
I would add a mention about the difficult introduced by microservices in the transaction management.
Think about an aborted paiment and the product stock control.
Surely this problem needs a complete blog entry to talk about it.
I'm not talking about a problem in my design. It is a problem in Microservices architectures.
It is covered for example in "Microservices Patterns" book, "Managing transactions with SAGAS". This is because I mentioned it, to add more information to your entry. π
I'll review it. Thanks for the clarification. ππΌ
I canβt wait until everyone stops using these terms. And instead learns hexagonal architecture vs n-tier vs workflow architecture vs plethora of architectures out there.
Only the Linux kernel 1.2 was large enough and not modular to be called a monolith. If you modularize, by definition not a monolith!
Rest endpoints + gateway is not an API! Unless your modules can use more than http. When that happens, even grpc or dynamic linking, no. No longer do have a microservice. You have an API.
There are some great points made. But please apply them beyond the two naive
terms.
Wow. Thanks for sharing.
Excellent concepts and overall explanation, can't wait to next articles on the serie
Can you consider scenarios for distributed transactions across services?
Regards
Microservices are the end to traditional back-end architecture. Today it should be cloud-endpoints only.
There are errors in this blog post, at a point that this is not serious.
Docker is not magically 'effective' because a team swaped to microservices.
This article promote microservices as if it was a magical response to the every day issues.
Note to readers: If your team is less than 20 tech people, and you are setting up a micro service architecture, then stop, micro services has many pros, so do monolithic.
Microservices is a hype by big-tech to drive people towards managed/serverless services, or dive deep into a rather unneeded complexity on your own, in most cases you need SOA, clean design in general pays off big time, but this increases in complexity exponentially. Hopefully the world would understand one day what is a hype
Π‘ongratulations π₯³! Your article hit the top posts for the week - dev.to/fruntend/top-10-posts-for-f...
Keep it up π
If we have databases for each service and one service handles users register and authentication, how does we duplicate this user information in all the database to give the permission across all services