---

Summary

This work proposes using eBPF to move application-defined storage functions deeper into the kernel I/O path. The approach targets fast NVMe devices where kernel overhead is a large share of access latency, while retaining important file-system properties such as safety and block-to-file-offset translation.

Bibtex

@inproceedings{zhong:bpf-storage,
  author = {Yuhong Zhong and Hongyi Wang and Yu Jian Wu and Asaf Cidon and Ryan Stutsman and Amy Tai and Junfeng Yang},
  title = {{BPF for Storage: An Exokernel-Inspired Approach}},
  booktitle = {Workshop on Hot Topics in Operating Systems},
  series = {HotOS '21},
  year = {2021},
  pages = {128--135},
  publisher = {ACM},
  url = {https://doi.org/10.1145/3458336.3465290},
  doi = {10.1145/3458336.3465290},
}

Generously Sponsored By

---