End-to-End Performance Analysis of Learning-enabled Systems
We propose a performance analysis tool for learning-enabled systems that allows operators to uncover potential performance issues before deploying DNNs in their systems. The tools that exist for this purpose require operators to faithfully model all components (a white-box approach) or do inefficient black-box local search. We propose a gray-box alternative, which eliminates the need to precisely model all the system's components. Our approach is faster and finds substantially worse scenarios compared to prior work. We show that a state-of-the-art learning-enabled traffic engineering pipeline can underperform the optimal by 6× --- a much higher number compared to what the authors found.
Authors
- Santiago Segarra (ORCID: https://orcid.org/0000-0002-8408-9633)
- Michael Schapira (ORCID: https://orcid.org/0000-0002-9336-8351)
- Ramesh Govindan
- Pooria Namyar (ORCID: https://orcid.org/0000-0001-7704-5440)
- Behnaz Arzani (ORCID: https://orcid.org/0000-0003-2234-3250)
- Ryan Beckett (ORCID: https://orcid.org/0000-0001-7844-2026)
- Siva Kesava Reddy Kakarla (ORCID: https://orcid.org/0009-0006-2694-4685)
Institutions
- University of Southern California (US)
- Hebrew University of Jerusalem (IL)
- California Southern University (US)
- Microsoft Research (United Kingdom) (GB)
- Rice University (US)
Publication Details
- Published
- 2024-11-11
- DOI
- https://doi.org/10.1145/3696348.3696875
- Citations
- 1
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.21
Funders
- National Science Foundation
- Israel Science Foundation