A typical microservice exposes its functionality as REST or gRPC interface and communicates with other services via a messaging framework, usually the publisher/subscriber way. The implementation is based on a programming framework like Spring Boot with Tomcat or Netty server implementation. The memory footprint is then around 200-300 MB, and the startup time is in seconds. These lead to limitations in local development, where a developer may not be able to start all necessary services, and also in production, where the cloud providers charge more for additional memory.
The thesis goal is to propose solutions to cut down memory footprint. Possible ways of doing so cover: joining more microservices into a single deployment, use of other Java VM, e.g., use of other implementations of java development kit or compilation in native code with GraalVM, and optimizing the programming framework itself, e.g., optimization of direct buffers and dependency injection.
This thesis is done in cooperation with GoodData.