A machine learning approach to service discovery for microservice architectures

Installation :

Prerequisites

• Java 8
• Maven 3
• Python 3.7
  • Modules:
    • Flask@1.1.1
    • joblib@0.14.1
    • Keras@2.3.1
    • numpy@1.19.1
    • pandas@1.0.1
    • pyzip@0.2.0
    • scikit-learn@0.22.2
    • tensorflow@1.15.0
• Mysql:
  • Databases:
    • num_service_monitoring.
      • Username: num_service_monitoring
      • Password: num_service_monitoring
      • Port 3306
    • numismatic_service.
      • Username: numismatic_service
      • Password: numismatic_service
      • Port 3306
    • user_service.
      • Username: user_service
      • Password: user_service
      • Port 3306
    • coin_service.
      • Username: coin_service
      • Password: coin_service
      • Port 3306

Build microservices

mvn -f ./discovery-client-module/pom.xml clean install
mvn -f ./api-gateway/pom.xml clean install
mvn -f ./numismatic-service/pom.xml clean install
mvn -f ./coin-service/pom.xml clean install
mvn -f ./service-discovery/pom.xml clean install
mvn -f ./service-monitoring/pom.xml clean install
mvn -f ./user-service/pom.xml clean install
mvn -f ./stub-service/pom.xml clean install
mvn -f ./auth-service/pom.xml clean install

App configuration

In config.yml file it's possible to configure the application (service discovery location, machine learning strategies and so on). You will find more details into config.yml file.

Start application

Run microservices

For each microservice, go to the target directory and execute the jar file.

For instance, for numismatic service:

cd numismatic-service/target
java -jar -Dspring.profiles.active=profile1 numismatic-service-0.0.1-SNAPSHOT.jar

You can specify different profile (from 1 to 5) in order to run multiple instances of the same microservice that have different response delays over time.

Run machine learning engine

cd ml-engine
python3 ./app.py

Run client

python3 client.py > client.log &

Author

This code was developed by Marco De Toma (https://github.com/detomarco)