Case Study - CodePix

About

• A solution to simulate value transfers between fictional banks through keys (email, cpf);
• We’ll simulate several banks and bank accounts that have an Pix key assigned;
• Each bank account will can to register your Pix key
• An bank account will can realise a transfer to the other account at other bank using an Pix key from the destination account;
• An transfer cannot be lost even that: CodePix system is down;
• An transfer cannot be lost even that: An destinantion Bank system is down;

About Banks

• A Bank will be a microservice with limited functions to register new account and Pix keys, as well as transfers of values;
• We’ll use the same application to simulate several banks, changing only colors, name and code.
• Nest.js in the backend
• Next.js as frontend

Idea

About CodePix

• The microservice CodePix will be responsible for intermediate the banks transfers
• Will receive a transaction of transfer
• Will forward to transaction from destination bank (Status: “pending”)
• Receive a confirmation from destionation bank (Status: “confirmed”)
• Send confirmation to the origin bank informing when destination bank processed
• Receive confirmation from origin bank that transaction was processed (Status: “completed”)
• Mark the transaction as complete (Status: “completed”)

Register and consulting a key Pix

• Consulte if key Pix exists and return if exists or not
• If not exists, we can create on sending a request to CodePix
• Key are storage in the bank & CodePix

Process dynamic

CodePix:

1. Register a transaction
2. Change status to: ‘Confirmed’
3. Informs origin bank that a transaction was confirmed by the destination bank with status: “confirmed
4. Finalize the transaction changing status to: “completed”

Main Challenges

• Fast and efficient communication
• Instant creation and query of keys (Synchronous)
• Guarantee that no transaction is lost even if any of the 3 systems are down (Asynchronous)

For this will use gRPC, a frameworkd that use HTTP2 and protocol buffers

For asynchronous comunication will use APACHE Kafka, to process data

CodePix

• Will be able to act as a gRPC server
• Consume and publish messages in Apache Kafka
• Both operations must be performed simultaneously when executing the service
• Work with a design focused on solving the domain problem
• Leave the technical complexity to the “application layer”, responsible for the gRPC and Kafka server
• Flexible for the implementation of other communication formats such as API Rest, CLI clients, etc. WITHOUT changing any other application components or the domain model.

CodePix structure and layers

Resources

• Docker
• Golang
• Apache Kafka
• Postgres