{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,10,8]],"date-time":"2025-10-08T18:10:53Z","timestamp":1759947053294,"version":"build-2065373602"},"reference-count":56,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,9,2]],"date-time":"2025-09-02T00:00:00Z","timestamp":1756771200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,9,2]],"date-time":"2025-09-02T00:00:00Z","timestamp":1756771200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["92267105"],"award-info":[{"award-number":["92267105"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,9,2]]},"DOI":"10.1109\/cluster59342.2025.11186468","type":"proceedings-article","created":{"date-parts":[[2025,10,7]],"date-time":"2025-10-07T17:35:09Z","timestamp":1759858509000},"page":"1-12","source":"Crossref","is-referenced-by-count":0,"title":["Scalable and Fast Inference Serving via Hybrid Communication Scheduling on Heterogeneous Networks"],"prefix":"10.1109","author":[{"given":"Gonglong","family":"Chen","sequence":"first","affiliation":[{"name":"Shenzhen Institutes of Advanced Technology, CAS,Shenzhen,China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jiamei","family":"Lv","sequence":"additional","affiliation":[{"name":"Zhejiang University,Hangzhou,China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Kejiang","family":"Ye","sequence":"additional","affiliation":[{"name":"Shenzhen Institutes of Advanced Technology, CAS,Shenzhen,China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Tao","family":"Gu","sequence":"additional","affiliation":[{"name":"Macquarie University,Sydney,Australia"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Chengzhong","family":"Xu","sequence":"additional","affiliation":[{"name":"University of Macau,Macau SAR,China"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"year":"2023","author":"Corporation","journal-title":"Worldwide ai and generative ai spending guide","key":"ref1"},{"key":"ref2","article-title":"Deepseek-r1: Incentivizing reasoning capability in 11 ms via reinforcement learning","author":"Guo","year":"2025","journal-title":"arXiv preprint"},{"year":"2023","author":"Guardian","journal-title":"Chatgpt reaches 100 million users two months after launch","key":"ref3"},{"volume-title":"Proceedings of USENIX OSDI","author":"Zhong","article-title":"Distserve: disaggregating prefill and decoding for goodput-optimized large language model serving","key":"ref4"},{"volume-title":"Proceedings of USENIX OSDI","author":"Agrawal","article-title":"Taming Throughput-Latency tradeoff in LLM inference with Sarathi-Serve","key":"ref5"},{"key":"ref6","article-title":"Fast distributed inference serving for large language models","author":"Wu","year":"2024","journal-title":"arXiv preprint"},{"volume-title":"Proceedings of ACM\/IEEE ISCA","author":"Patel","article-title":"Splitwise: Efficient generative 11 m inference using phase splitting","key":"ref7"},{"volume-title":"Proceedings of USENIX OSDI","author":"Yu","article-title":"Orca: A distributed serving system for Transformer-Based generative models","key":"ref8"},{"doi-asserted-by":"publisher","key":"ref9","DOI":"10.1145\/3600006.3613165"},{"year":"2025","journal-title":"NVIDIA","article-title":"Nvidia connectx infiniband adapters","key":"ref10"},{"doi-asserted-by":"publisher","key":"ref11","DOI":"10.1145\/3452296.3472899"},{"volume-title":"Proceedings of USENIX NSDI","author":"Lao","article-title":"ATP: In-network aggregation for multi-tenant learning","key":"ref12"},{"volume-title":"Proceedings of USENIX NSDI","author":"Sapio","article-title":"Scaling distributed machine learning with In-Network aggregation","key":"ref13"},{"doi-asserted-by":"publisher","key":"ref14","DOI":"10.1109\/ICNP59255.2023.10355636"},{"doi-asserted-by":"publisher","key":"ref15","DOI":"10.1145\/3603269.3604835"},{"doi-asserted-by":"publisher","key":"ref16","DOI":"10.1145\/3617232.3624863"},{"doi-asserted-by":"publisher","key":"ref17","DOI":"10.1109\/IWQoS61813.2024.10682850"},{"doi-asserted-by":"publisher","key":"ref18","DOI":"10.1145\/3575693.3575708"},{"doi-asserted-by":"publisher","key":"ref19","DOI":"10.1109\/INFOCOM53939.2023.10228956"},{"doi-asserted-by":"publisher","key":"ref20","DOI":"10.1145\/3582016.3582037"},{"doi-asserted-by":"publisher","key":"ref21","DOI":"10.1109\/SC41404.2022.00051"},{"key":"ref22","article-title":"In-network aggregation for shared machine learning clusters","volume-title":"Proceedings of Machine Learning and Systems","volume":"3","author":"Gebara"},{"year":"2025","journal-title":"Nvidia a100","key":"ref23"},{"year":"2023","journal-title":"Sharegpt teams","key":"ref24"},{"doi-asserted-by":"publisher","key":"ref25","DOI":"10.18653\/v1\/2024.acl-long.172"},{"key":"ref26","article-title":"Opt: Open pre-trained transformer language models","author":"Zhang","year":"2022","journal-title":"arXiv preprint"},{"volume-title":"Proceedings of NIPS","author":"Vaswani","article-title":"Attention is all you need","key":"ref27"},{"volume-title":"Proceedings of NIPS","author":"Brown","article-title":"Language models are few-shot learners","key":"ref28"},{"volume-title":"Proceedings of NIPS","author":"Ouyang","article-title":"Training language models to follow instructions with human feedback","key":"ref29"},{"key":"ref30","article-title":"Scaling laws for neural language models","author":"Kaplan","year":"2020","journal-title":"arXiv preprint"},{"key":"ref31","article-title":"The economic implications of large language model selection on earnings and return on investment: A decision theoretic model","author":"Xex\u00e9o","year":"2024","journal-title":"arXiv preprint"},{"doi-asserted-by":"publisher","key":"ref32","DOI":"10.1145\/3560815"},{"key":"ref33","article-title":"Flash communication: Reducing tensor parallelization bottleneck for fast large language model inference","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref34","article-title":"Efficient deployment of large language models on resource-constrained devices","author":"Yao","year":"2025","journal-title":"arXiv preprint"},{"key":"ref35","article-title":"Kvdirect: Distributed disaggregated llm inference","author":"Chen","year":"2024","journal-title":"arXiv preprint"},{"doi-asserted-by":"publisher","key":"ref36","DOI":"10.1145\/3651890.3672233"},{"volume-title":"Proceedings of USENIX FAST","author":"Qin","article-title":"Mooncake: Trading more storage for less computation-a KVCache-centric architecture for serving LLM chatbot","key":"ref37"},{"year":"2025","author":"Zhao","journal-title":"Deepep: an efficient expert-parallel communication library","key":"ref38"},{"volume-title":"Proceedings of USENIX OSDI","author":"Gujarati","article-title":"Serving DNNs like clockwork: Performance predictability from the bottom up","key":"ref39"},{"volume-title":"Proceedings of USENIX OSDI","author":"Li","article-title":"AlpaServe: Statistical multiplexing with model parallelism for deep learning serving","key":"ref40"},{"key":"ref41","article-title":"Megatron-lm: Training multi-billion parameter language models using model parallelism","author":"Shoeybi","year":"2019","journal-title":"arXiv preprint"},{"year":"2024","journal-title":"Intel","article-title":"Intel intelligent fabric processors","key":"ref42"},{"key":"ref43","article-title":"Tiara: A scalable and efficient hardware acceleration architecture for stateful layer-4 load balancing","volume-title":"Proc. of USENIX NSDI","author":"Zeng","year":"2022"},{"doi-asserted-by":"publisher","key":"ref44","DOI":"10.1109\/TPDS.2021.3094364"},{"doi-asserted-by":"publisher","key":"ref45","DOI":"10.1109\/TPAMI.2023.3237667"},{"year":"2025","author":"Shengyu","journal-title":"Swifttransformer: High performance transformer implementation in c++","key":"ref46"},{"year":"2025","journal-title":"Nvidia dcgm","key":"ref47"},{"year":"2024","author":"Lin","journal-title":"Apex: An extensible and dynamism-aware simulator for automated parallel execution in 11 m serving","key":"ref48"},{"volume-title":"Proceedings of USENIX OSDI","author":"Li","article-title":"AlpaServe: Statistical multiplexing with model parallelism for deep learning serving","key":"ref49"},{"volume-title":"Proceedings of ACM ASPLOS","author":"Xu","article-title":"Fast on-device 11 m inference with npus","key":"ref50"},{"year":"2019","author":"Corporation","journal-title":"Fastertransformer","key":"ref51"},{"doi-asserted-by":"publisher","key":"ref52","DOI":"10.1109\/ICNP61940.2024.10858501"},{"doi-asserted-by":"publisher","key":"ref53","DOI":"10.1109\/ICNP59255.2023.10355574"},{"doi-asserted-by":"publisher","key":"ref54","DOI":"10.1109\/IPCCC59175.2023.10253874"},{"doi-asserted-by":"publisher","key":"ref55","DOI":"10.1109\/INFOCOM52122.2024.10621230"},{"volume-title":"Proceedings of USENIX NSDI","author":"Zhao","article-title":"NetRPC: Enabling In-Network computation in remote procedure calls","key":"ref56"}],"event":{"name":"2025 IEEE International Conference on Cluster Computing (CLUSTER)","start":{"date-parts":[[2025,9,2]]},"location":"United Kingdom","end":{"date-parts":[[2025,9,5]]}},"container-title":["2025 IEEE International Conference on Cluster Computing (CLUSTER)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11186399\/11186452\/11186468.pdf?arnumber=11186468","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,8]],"date-time":"2025-10-08T17:36:45Z","timestamp":1759945005000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11186468\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,9,2]]},"references-count":56,"URL":"https:\/\/doi.org\/10.1109\/cluster59342.2025.11186468","relation":{},"subject":[],"published":{"date-parts":[[2025,9,2]]}}}