MicroSuggest: Kernel-Aware Microservice Decomposition

Harsh Borse, Utkalika Satapathy, Mainack Mandal, Bivas Mitra

Abstract

Microservice decomposition typically emphasizes logical or domain-driven boundaries, often overlooking performance bottlenecks from low-level system interactions. We present a system call-aware decomposition method that identifies and separates functions likely to interfere at the kernel level. By defining a collision score based on system call frequency and type, and using a fine-tuned Large Language Model to statically predict syscall behavior, we construct a function interaction graph for clustering. Evaluation on Python-based monoliths shows up to 30% latency reduction and improved scalability compared to traditional approaches, demonstrating the value of kernel-informed microservice design.

Bibtex

@inproceedings{borse2025microsuggest,
  title={MicroSuggest: Kernel-Aware Microservice Decomposition},
  author={Borse, Harsh and Satpathy, Utkalika and Mondal, Mainack and Mitra, Bivas},
  booktitle={International Conference on Big Data Analytics and Knowledge Discovery},
  pages={302--308},
  year={2025},
  organization={Springer}
}

Cite

Borse, H., Satpathy, U., Mondal, M., & Mitra, B. (2025, August). MicroSuggest: Kernel-Aware Microservice Decomposition. In International Conference on Big Data Analytics and Knowledge Discovery (pp. 302-308). Cham: Springer Nature Switzerland.

← Back to all publications