GPUs, CPUs, and… NICs: Rethinking the Network’s Role in Serving Complex AI Pipelines
Mike Wong, Ulysses Butler, Emma Farkash, Praveen Tammana, Anirudh Sivaraman, Ravi Netravali · 2025
Read GPUs, CPUs, and… NICs: Rethinking the Network’s Role in Serving Complex AI Pipelines by Mike Wong, Ulysses Butler, Emma Farkash, Praveen Tammana, Anirudh Sivaraman, Ravi Netravali (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-15712.txt.
Description
Mike Wong, Ulysses Butler, Emma Farkash, Praveen Tammana, Anirudh Sivaraman, and Ravi Netravali argue that the network card deserves a seat at the AI serving table. Modern AI pipelines chain many stages together and run them across machines, which makes network delays a fact of life and piles up resource overhead. The authors flip the framing: instead of treating the network as a bottleneck, they ask what it can absorb.
Their claim is that the heavy data processing inside AI pipelines matches the strengths of packet processing hardware. Much of it can be offloaded onto SmartNICs, the programmable network cards at each server. The paper explores the challenges and openings of that offload and lays out a research agenda for weaving network hardware into AI pipelines. The aim is to cut the overheads that now tax GPUs and CPUs.
Catalog record
| CREATOR | Mike Wong, Ulysses Butler, Emma Farkash, Praveen Tammana, Anirudh Sivaraman, Ravi Netravali |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-15712