This research leverages Large Language Models to generate synthetic microservice call graphs that closely mimic real-world system behavior.

• Addresses the challenge of limited access to real-world traces for testing and optimization
• Trains LLMs to capture complex hierarchical structures and constraints in microservice interactions
• Provides a novel solution for engineers to test systems without sensitive production data
• Enables more accurate capacity planning and resource management

For engineering teams, this approach offers a practical way to evaluate system performance and reliability without requiring access to proprietary or sensitive operational data.

Large Language Models as Realistic Microservice Trace Generators