Welcome to Habitus Project
- Habitus is a project for encouraging the power of habits among users 💪
- It is developed using microservices architecture 🌈
Check out all the other diagrams (e.g Deployment view diagram and Data View Model) in the Diagram and Documentation repository.
- User Authentication
- User Authorization
- Zero trust
- Limit exposure: Using private networks for the microservices, message queue and LDAP.
- Data Encryption
- Actions statistics: Using GCP one can have a registry of the actions users perform.
- Interoperability between microservices using API Gateway orchestration and reverse proxies.
- REST is used for communication between API Gateway and microservices. GraphQL is used between frontend and Gateway.
- There's also a SOAP service for communication between the system and others that know the IP of the reverse proxy connected to that service.
- Each microservice uses a specific communication tool for communicating with their asociated database.
- Microservices architecture
- Implementation automation using Kubernetes.
- Horizontal Scalability: Load balancing services.
- Event based async communication between microservices.
- Testing: Load and stress testing done with Jmeter.
Testing was done with Jmeter. An example of a test case is creating a large number of logins (which thus needs to access the users_ms microservice and the LDAP server).
Only one computing instance
| # Users | Response time (ms) | Throughput (trans/min) |
|---|---|---|
| 1 | 632 | 36.8 |
| 50 | 890 | 1587.3 |
| 100 | 1403 | 2496.9 |
| 105 | 1564 | 2457.1 |
| 110 | 4930 | 1113.0 |
| 114 | 16700 | 383.1 |
| 115 | 58959 | 115.1 |
| 125 | - | - |
The knee-point is obtained at 115 concurrent logins when the first error is identified.
Scaled to 3 replicas.
| # Users | Response time (ms) | Throughput (trans/min) |
|---|---|---|
| 1 | 541 | 38.9 |
| 50 | 889 | 1588.1 |
| 100 | 1280 | 2631.6 |
| 120 | 1523 | 2853.7 |
| 130 | 1730 | 2857.1 |
| 135 | 18470 | 416.0 |
| 136 | 104940 | 77.0 |
| Here the knee-point is obtained at 136 concurrent logins where an error is identified. This replication technique improved the response time in 18.42%. |