{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,8]],"date-time":"2026-09-08T05:12:05Z","timestamp":1788844325428,"version":"build-2803163510"},"reference-count":39,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,7]]},"DOI":"10.1109\/icws72778.2026.00182","type":"proceedings-article","created":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T19:18:56Z","timestamp":1788808736000},"page":"1490-1500","source":"Crossref","is-referenced-by-count":0,"title":["EcoFlow: Cost-Aware LLM Serving on Heterogeneous GPUs via Min-Cost Max-Flow"],"prefix":"10.1109","author":[{"given":"Fengkai","family":"Jiang","sequence":"first","affiliation":[{"name":"Guangdong University of Technology","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Fulu","family":"Kou","sequence":"additional","affiliation":[{"name":"Guangdong University of Technology","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jinquan","family":"Zhang","sequence":"additional","affiliation":[{"name":"Guangdong University of Technology","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Rui","family":"Yang","sequence":"additional","affiliation":[{"name":"Guangdong University of Technology","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shujie","family":"Pang","sequence":"additional","affiliation":[{"name":"Guangdong University of Technology","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/ICWS62655.2024.00098"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/ICWS67624.2025.00046"},{"key":"ref3","article-title":"OpenAI GPT-5 system card","author":"Singh","year":"2026","journal-title":"arXiv preprint"},{"key":"ref4","year":"2025","journal-title":"Gemini 3: A family of highly capable multimodal models"},{"key":"ref5","year":"2025","journal-title":"DeepSeek-V4: Towards highly efficient million-token context intelligence"},{"key":"ref6","article-title":"Qwen3 technical report","author":"Yang","year":"2025","journal-title":"arXiv preprint"},{"key":"ref7","article-title":"The Llama 3 herd of models","author":"Grattafiori","year":"2024","journal-title":"arXiv preprint"},{"key":"ref8","year":"2025","journal-title":"The Llama 4 herd: The beginning of a new era of natively multimodal AI"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/TSC.2022.3196915"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1007\/s00607-023-01153-1"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.4018\/JOEUC.20210501.oa4"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1145\/3669940.3707215"},{"key":"ref13","article-title":"Language models are unsupervised multitask learners","author":"Radford","year":"2019","journal-title":"OpenAI Blog"},{"key":"ref14","first-page":"1877","article-title":"Language models are few-shot learners","volume-title":"Proc. NeurIPS","volume":"33","author":"Brown","year":"2020"},{"key":"ref15","article-title":"Demystifying cost-efficiency in LLM serving over heterogeneous GPUs","volume-title":"Proc. ICML","author":"Jiang","year":"2025"},{"key":"ref16","first-page":"521","article-title":"Orca: A distributed serving system for transformer-based generative models","volume-title":"Proc. USENIX OSDI","author":"Yu","year":"2022"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1145\/3600006.3613165"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA59077.2024.00019"},{"key":"ref19","article-title":"DistServe: Disaggregating prefill and decoding for goodput-optimized large language model serving","volume-title":"Proc. USENIX OSDI","author":"Zhong","year":"2024"},{"key":"ref20","article-title":"Mooncake: Trading more storage for less computation-A KVCache-centric architecture for serving LLM chatbot","volume-title":"Proc. 23rd USENIX Conf. File Storage Technol. (FAST)","author":"Qin","year":"2025"},{"key":"ref21","article-title":"FlexGen: High-throughput generative inference of large language models with a single GPU","volume-title":"Proc. ICML","author":"Sheng","year":"2023"},{"key":"ref22","article-title":"GPipe: Efficient training of giant neural networks using pipeline parallelism","volume-title":"Proc. NeurIPS","author":"Huang","year":"2019"},{"key":"ref23","article-title":"Megatron-LM: Training multi-billion parameter language models using model parallelism","volume-title":"Proc. Int. Conf. High Perform. Comput., Netw., Storage Anal. (SC)","author":"Shoeybi","year":"2019"},{"key":"ref24","article-title":"HexGen: Generative inference of LLM over heterogeneous environment","volume-title":"Proc. ICML","author":"Jiang","year":"2024"},{"key":"ref25","article-title":"HexGen-2: Disaggregated generative inference of LLMs in heterogeneous environment","volume-title":"Proc. 13th Int. Conf. Learn. Representations (ICLR)","author":"Jiang","year":"2025"},{"key":"ref26","article-title":"TriInfer: Hybrid EPD disaggregation for efficient multimodal large language model inference","volume-title":"Proc. MLSys","author":"Dong","year":"2026"},{"key":"ref27","article-title":"AlpaServe: Statistical multiplexing with model parallelism for deep learning serving","volume-title":"Proc. USENIX OSDI","author":"Li","year":"2023"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1145\/3797823.3797828"},{"key":"ref29","article-title":"SkyPilot: An intercloud broker for sky computing","volume-title":"Proc. USENIX NSDI","author":"Yang","year":"2023"},{"key":"ref30","article-title":"Coral: Cost-efficient multi-LLM serving over heterogeneous cloud GPUs","author":"Mei","year":"2026","journal-title":"arXiv preprint"},{"key":"ref31","article-title":"M\u00e9lange: Cost efficient LLM serving by exploiting GPU heterogeneity","author":"Griggs","year":"2024","journal-title":"arXiv preprint"},{"key":"ref32","article-title":"Scalable joint resource allocation for SLOconstrained LLM inference in heterogeneous GPU clouds","author":"Cheng","year":"2026","journal-title":"arXiv preprint"},{"key":"ref33","article-title":"BOute: Cost-efficient LLM serving with heterogeneous LLMs and GPUs via multi-objective Bayesian optimization","volume-title":"Proc. Mach. Learn. Syst. (MLSys)","author":"Jiang","year":"2026"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/3627703.3629578"},{"key":"ref35","article-title":"Taming throughput-latency tradeoff in LLM inference with Sarathi-Serve","volume-title":"Proc. USENIX OSDI","author":"Agrawal","year":"2024"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/ITC3249928.2020.00016"},{"key":"ref37","article-title":"ModServe: Modality- and stage-aware resource disaggregation for scalable multimodal model serving","volume-title":"Proc. ACM SoCC","author":"Qiu","year":"2025"},{"key":"ref38","year":"2026","journal-title":"Compute Engine GPU pricing"},{"key":"ref39","article-title":"BEAM: Joint resource-power optimization for energy-efficient LLM inference under SLO constraints","volume-title":"Proc. MLSys","author":"Lee","year":"2026"}],"event":{"name":"2026 IEEE International Conference on Web Services (ICWS)","location":"Sydney, Australia","start":{"date-parts":[[2026,7,13]]},"end":{"date-parts":[[2026,7,18]]}},"container-title":["2026 IEEE International Conference on Web Services (ICWS)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11670018\/11670022\/11670094.pdf?arnumber=11670094","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,8]],"date-time":"2026-09-08T04:53:03Z","timestamp":1788843183000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11670094\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7]]},"references-count":39,"URL":"https:\/\/doi.org\/10.1109\/icws72778.2026.00182","relation":{},"subject":[],"published":{"date-parts":[[2026,7]]}}}