Skip to content
master
Go to file
Code

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 

README.md

Metis: Interpreting Deep Learning-Based Networking Systems

Metis is an integrated explainer to provide post-hoc interpretation for different types of Deep Learning (DL)-based networked systems. We refer the readers to the technical report for more details.

In the current stage, we provide the interpretation methods and implementations for three DL-based networking systems:

We further provide four use cases of Metis:

  • Metis helps network operators to redesign the DNN structure of Pensieve with a quality of experience (QoE) improvement by 5.1% on average (case-design).
  • Metis debugs the DNN in Pensieve and improves the average QoE by up to 4% with only decision trees (case-debug).
  • Metis enables a lightweight DL-based flow scheduler (AuTO) and a lightweight Pensieve with shorter decision latency by 27x and lower resource consumption by up to 156x (case-deploy).
  • Metis helps network operators to adjust the routing paths of a DL-based routing optimizer (RouteNet) when ad-hoc adjustments are needed (case-adjust).

The running scripts for interpretation methods and use cases could be found in respective directories. Currently we are still working on documentating and refactoring the repository. Other codes will be released soon. Please stay tuned!

For any questions, please post an issue or send an email to zilim@ieee.org.

Citation

@inproceedings{meng2020interpreting,
  title={Interpreting Deep Learning-Based Networking Systems},
  author={Meng, Zili and Wang, Minhu and Bai, Jiasong and Xu, Mingwei and Mao, Hongzi and Hu, Hongxin},
  booktitle={Proc. ACM SIGCOMM},
  year={2020}
}

About

Interpreting Deep Learning-Based Networking Systems (SIGCOMM 2020)

Resources

License

Releases

No releases published

Contributors 4

  •  
  •  
  •  
  •  
You can’t perform that action at this time.