Modular and Distributed IDE
summary
The paper generates deployable language-service microservices from Xtext/EMF specs, implements them with Quarkus, containers them with Docker and Kubernetes, and drives deployment with a FeatureIDE feature model. Measurements show completion stays ~5 ms but grows when remote, references go from <200 ms to nearly double, rename reaches ~7 s, and the NabLab compiler hits ~15 minutes; distributing yields 16–25% faster rename and 9–21% faster compiler but worsens light services. The main cost is model loading overhead when using stateless microservices; modularization and comms add little.
AI-written summary · about
chapters (12)
- 0:00Modular and distributed IDEs
- 1:28Monolithic language servers limits
- 2:10Motivation for distributed services
- 4:21Modularizing and deploying IDE services
- 5:26IDE microservicesation approach
- 6:53Protocol specification language
- 8:43Evaluation setup and architectures
- 9:27Monolithic vs microservices performance
- 11:16Performance differences root causes
- 12:27Statelessness and model loading impact
- 14:13Findings and benefits summary
- 14:57Open research challenges
related talks
36mA day in life of a Quarkus developer
37mQuarkus in Action…in Action!
30mAbout the Best Development Tooling in the World
55mWhy You Want to Code With Quarkus in 2023
35mHow we brought down our build times. Modularization impact on build speeds and metrics
50mMicroservices for building an IDE
17mDeveloper Experience Q&A
45mContaining Your Development – Why Cloud IDEs Are the Future