Metraipolis › archive › dc_6bb18b520cd0

RingStitch: Demand-Aware Optical Stitching for Fragmented TPU Clusters — arXiv:2609.37627

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.37627
License (as archived): CC0-1.0
Archived: 2026-10-07T17:59:39.369Z
SHA-256 of the archived text: ff213720f9f68f6fc6683512cd5d30c9fe14472b45f5f8a07ed73268d90c0446
Stable URL of this page: https://metraipolis.com/documento/dc_6bb18b520cd0

How to cite

RingStitch: Demand-Aware Optical Stitching for Fragmented TPU Clusters — arXiv:2609.37627 — arXiv (en), https://arxiv.org/abs/2609.37627 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6bb18b520cd0 (Metraipolis archive dc_6bb18b520cd0, sha256 ff213720f9f68f6fc6683512cd5d30c9fe14472b45f5f8a07ed73268d90c0446)

BibTeX

@misc{dc6bb18b520cd0,
  title = {RingStitch: Demand-Aware Optical Stitching for Fragmented TPU Clusters — arXiv:2609.37627},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.37627},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6bb18b520cd0; sha256 ff213720f9f68f6fc6683512cd5d30c9fe14472b45f5f8a07ed73268d90c0446},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.37627. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large-scale AI training clusters increasingly use optical circuit switching (OCS) to reconfigure rack-level interconnects and create elastic accelerator slices. In multi-tenant TPU-style clusters, however, small and medium jobs often leave partial free capacity stranded inside racks. Although the aggregate free capacity may be sufficient for a new job, it cannot be used by local placement or coarse full-rack stitching. This paper presents RingStitch, an OCS-based defragmentation scheduler that turns fragmented rack capacity into schedulable resources. RingStitch follows a local-first policy, stitches compact cross-rack fragments only when needed, and orders the selected racks into a low-cost logical ring. Simulations on a TPU 8t-like SuperPod model show that RingStitch improves schedulability over local and full-rack baselines.

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-30 in cs.NI).
Categorie arXiv: Networking and Internet Architecture (cs.NI).
Pagina del lavoro: https://arxiv.org/abs/2609.37627 — 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 2026-10-07T17:33:10.178Z dalla pagina https://arxiv.org/catchup/cs.NI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).