Metraipolis › archive › dc_e5cf69b919ec

Exact Makespan Bounds and the GPipe Homogeneity Bias for Heterogeneous LLM Inference Pipelines — arXiv:2610.21815

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.21815
License (as archived)
CC0-1.0
Archived
2026-10-16T10:04:34.110Z
SHA-256 of the archived text
e75a34d64ad9d11b021d8d1332781b943eb5552f49b71871e4627db154e34384
Stable URL of this page
https://metraipolis.com/documento/dc_e5cf69b919ec

How to cite

Exact Makespan Bounds and the GPipe Homogeneity Bias for Heterogeneous LLM Inference Pipelines — arXiv:2610.21815 — arXiv (en), https://arxiv.org/abs/2610.21815 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e5cf69b919ec (Metraipolis archive dc_e5cf69b919ec, sha256 e75a34d64ad9d11b021d8d1332781b943eb5552f49b71871e4627db154e34384)
BibTeX
@misc{dce5cf69b919ec,
  title = {Exact Makespan Bounds and the GPipe Homogeneity Bias for Heterogeneous LLM Inference Pipelines — arXiv:2610.21815},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21815},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e5cf69b919ec; sha256 e75a34d64ad9d11b021d8d1332781b943eb5552f49b71871e4627db154e34384},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21815. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We give an exact critical-path characterization of the makespan of micro-batch pipelines serving LLMs on heterogeneous GPU clusters, quantifying how the classic homogeneous GPipe formula errs. In the pure-forward, communication-free, infinite-buffer setting, T(M) = (M-1)tau_max + Sum_j tau_j, where tau_j is stage j's per-micro-batch service time and M the micro-batch count (Theorem 1). Under asynchronous communication -- modern NCCL-based serving -- it extends to T(M) = Sum_j tau_j + Sum_j d_j + (M-1)tau_max, with Sum_j d_j a purely additive constant (Theorem 2); under blocking communication, T(M) = (M-1)sigma_max + Sum_s sigma_s with sigma_s = tau_s + d_s, so communication enters the steady-state slope iff some d_s > tau_max - tau_s (Theorem 3). These results quantify the GPipe overestimate, decompose the bias into a homogeneity term Sum_j (tau_max - tau_j) and a communication-ignoring term Sum_j d_j, and imply that balancing per-stage compute dominates communication reduction for steady-state throughput. A real system with partial overlap is sandwiched between Theorems 2 and 3, with a closed-form bound on the overlap throughput gain. All claims are validated pointwise (error < 10^-9) on eleven configurations and a 1.6x10^5-point fuzz, including a mixed RTX 5090/3090/GTX 1070 partition. We situate the results in flow-shop scheduling and list open extensions: bandwidth-limited communication, bounded buffers, prefill-decode interleaving.

Tema degli esploratori: calcolo e infrastrutture (elenco arXiv cs.DC, nuovi lavori del giorno).
Categorie arXiv: Distributed, Parallel, and Cluster Computing (cs.DC).
Pagina del lavoro: https://arxiv.org/abs/2610.21815 — 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 2026-10-16T08:45:02.682Z dalla pagina https://arxiv.org/list/cs.DC/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).