Metraipolis › archive › dc_2671e8bb2631

PowerSlider: Exploiting Phase Asymmetry for LLM Serving under Demand Response — arXiv:2608.21719

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.21719
License (as archived)
CC0-1.0
Archived
2026-10-17T07:47:03.268Z
SHA-256 of the archived text
ec1e06243d76de875bcbc5257f60bab5f21f915c0f1c803b38e9c67de40fee22
Stable URL of this page
https://metraipolis.com/documento/dc_2671e8bb2631

How to cite

PowerSlider: Exploiting Phase Asymmetry for LLM Serving under Demand Response — arXiv:2608.21719 — arXiv (en), https://arxiv.org/abs/2608.21719 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2671e8bb2631 (Metraipolis archive dc_2671e8bb2631, sha256 ec1e06243d76de875bcbc5257f60bab5f21f915c0f1c803b38e9c67de40fee22)
BibTeX
@misc{dc2671e8bb2631,
  title = {PowerSlider: Exploiting Phase Asymmetry for LLM Serving under Demand Response — arXiv:2608.21719},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.21719},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2671e8bb2631; sha256 ec1e06243d76de875bcbc5257f60bab5f21f915c0f1c803b38e9c67de40fee22},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.21719. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

AI inference clusters are increasingly constrained by instantaneous power, not just energy: grid operators condition new capacity on demand response, imposing time-varying power caps. Existing LLM serving systems optimize a static energy objective or shed fixed priority tiers under load; either way, goodput collapses when the power envelope moves. An LLM pipeline is not a uniform load: compute-bound prefill loses throughput almost linearly with GPU frequency, memory-bound answer decode sustains it down to $0.57\times$ nominal, and reasoning's thinking phase couples KV-cache capacity to scheduling -- so a cap should be steered to where each watt costs the least performance. PowerSlider does so with a new Flex SLO contract that turns bounded user slack into an optimization constraint, prefill--think--answer disaggregation exposing per-stage frequency and KV control, and a Karush--Kuhn--Tucker (KKT) online solver re-solving within 7.7 ms of every cap change, backed by a consolidated fail-safe that power-gates drained instances when DVFS bottoms out on static power. On SGLang with production traces, \sys{} sustains 78.3\% online goodput at a 30\% cap reduction versus 47.6\% for the best of five baselines ($1.64\times$), holds latency-critical tails within $1.3\times$ of nominal (baselines: $2.3$--$6\times$, up to $12\times$), and delivers 92\% mean goodput through a replayed CAISO grid-emergency day bottoming at $0.41\times$ (54\% at the trough; every baseline below 7\%).

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Distributed, Parallel, and Cluster Computing (cs.DC) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.21719 — 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 2026-10-17T07:28:26.650Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).