Metraipolis › archive › dc_4f0b69e4239f
MoE Expert Execution in Disaggregated LLM Serving with a High-Bandwidth ReRAM Near-Memory Architecture — arXiv:2608.13962
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.13962
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T02:28:06.561Z
- SHA-256 of the archived text
93a6acf69622d80fe8b74b2549e593a73cbb987b7283f5de6ac63ef7dc5ddd80- Stable URL of this page
https://metraipolis.com/documento/dc_4f0b69e4239f
How to cite
MoE Expert Execution in Disaggregated LLM Serving with a High-Bandwidth ReRAM Near-Memory Architecture — arXiv:2608.13962 — arXiv (en), https://arxiv.org/abs/2608.13962 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4f0b69e4239f (Metraipolis archive dc_4f0b69e4239f, sha256 93a6acf69622d80fe8b74b2549e593a73cbb987b7283f5de6ac63ef7dc5ddd80)
BibTeX
@misc{dc4f0b69e4239f,
title = {MoE Expert Execution in Disaggregated LLM Serving with a High-Bandwidth ReRAM Near-Memory Architecture — arXiv:2608.13962},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.13962},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4f0b69e4239f; sha256 93a6acf69622d80fe8b74b2549e593a73cbb987b7283f5de6ac63ef7dc5ddd80},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.13962. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Attention-FFN disaggregation maps LLM modules to specialized pools, creating an opening to keep Mixture-of-Experts (MoE) weights resident in a high-bandwidth FFN pool. Decode SLOs, however, cap the run-batch while sparse routing expands the activated-expert union, so weight traffic amortizes poorly and routing skew idles cold-expert resources. The FFN pool must therefore deliver weight-read bandwidth density under sparse unions and recover occupancy under skew without a global sharing fabric. We present a ReRAM near-memory architecture that keeps expert weights resident behind high-bandwidth local reads. The design factors actual MFU into ideal MFU and occupancy, recovers occupancy with bounded core-local multicast pooling, coactivation-aware placement, and load-aware fetch, and sizes each communication level from induced demand. A measured + modeled study on Qwen3.5-35B-A3B, Qwen3.5-397B-A17B, and GLM-5.2 shows that side-4 pooling raises occupancy from 0.328 to 0.519 and, at iso-peak compute, lowers per-token FFN-pool latency by 9.5x versus H20 with 20x lower weight-movement energy; an H20-attention + ReRAM-FFN system reduces decode TPOT by 1.25-4.0x, 2.4-10.3x, and 2.5-10.4x versus a homogeneous H20 pool. Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-08-17 in cs.AR). Categorie arXiv: Hardware Architecture (cs.AR). Pagina del lavoro: https://arxiv.org/abs/2608.13962 — 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 2026-10-19T02:21:05.702Z dalla pagina https://arxiv.org/catchup/cs.AR/2026-08-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).