Metraipolis › archive › dc_b48f5dc4e2fc

ARK: Avoiding Routing Collisions for KV Cache Transfer in Disaggregated LLM Inference — arXiv:2610.41160

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.41160
License (as archived)
CC0-1.0
Archived
2026-11-03T07:12:38.102Z
SHA-256 of the archived text
5ba5cd0389fed2084d083ad9587795e8d70ce082b5b76519a2b9e42750542f91
Stable URL of this page
https://metraipolis.com/documento/dc_b48f5dc4e2fc

How to cite

ARK: Avoiding Routing Collisions for KV Cache Transfer in Disaggregated LLM Inference — arXiv:2610.41160 — arXiv (en), https://arxiv.org/abs/2610.41160 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b48f5dc4e2fc (Metraipolis archive dc_b48f5dc4e2fc, sha256 5ba5cd0389fed2084d083ad9587795e8d70ce082b5b76519a2b9e42750542f91)
BibTeX
@misc{dcb48f5dc4e2fc,
  title = {ARK: Avoiding Routing Collisions for KV Cache Transfer in Disaggregated LLM Inference — arXiv:2610.41160},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.41160},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b48f5dc4e2fc; sha256 5ba5cd0389fed2084d083ad9587795e8d70ce082b5b76519a2b9e42750542f91},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.41160. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Disaggregated LLM inference separates the prefill and decode phases across GPU pools, generating massive KV-cache transfers. Because these transfers last hundreds of milliseconds to seconds, they behave as mega elephant flows that dominate link bandwidth utilization. In this regime, stateless ECMP can perform poorly: hash collisions may overload one spine link while leaving others idle, stretching transfer times by seconds. Yet this same persistence makes coordination practical. Since these flows are long-lived, even lightweight one-to-all coordination can be amortized over their lifetime. We present ARK, a distributed elephant-flow path reservation mechanism. ARK coordinates senders to choose source ports whose hashes map concurrent flows onto distinct spines, without requiring switch changes or receiver-side packet reordering. Packet-level RDMA simulations show that ARK reduces mean and P95 FCT by up to 27.3% and 34.0% under moderate load, and further reduces mean TTFT by up to 12.9%.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Networking and Internet Architecture (cs.NI) ; Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC).
Pagina del lavoro: https://arxiv.org/abs/2610.41160 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-03T07:09:31.847Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).