{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,11,6]],"date-time":"2025-11-06T19:04:44Z","timestamp":1762455884085,"version":"build-2065373602"},"reference-count":47,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"12","license":[{"start":{"date-parts":[[2025,12,1]],"date-time":"2025-12-01T00:00:00Z","timestamp":1764547200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,12,1]],"date-time":"2025-12-01T00:00:00Z","timestamp":1764547200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,12,1]],"date-time":"2025-12-01T00:00:00Z","timestamp":1764547200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"National Key R&D Program of China","award":["2022YFF0604500"],"award-info":[{"award-number":["2022YFF0604500"]}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62432004","62402268"],"award-info":[{"award-number":["62432004","62402268"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100020721","name":"Guoqiang Institute, Tsinghua University","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100020721","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Tsinghua University (AIR)-AsiaInfo Technologies"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. on Mobile Comput."],"published-print":{"date-parts":[[2025,12]]},"DOI":"10.1109\/tmc.2025.3592647","type":"journal-article","created":{"date-parts":[[2025,7,24]],"date-time":"2025-07-24T17:57:58Z","timestamp":1753379878000},"page":"13309-13321","source":"Crossref","is-referenced-by-count":0,"title":["Squeezer: Efficient Multi-DNN Inference for Edge Video Analytics via Cross-Model Scheduling"],"prefix":"10.1109","volume":"24","author":[{"ORCID":"https:\/\/orcid.org\/0009-0003-3223-5744","authenticated-orcid":false,"given":"Xiang","family":"Wang","sequence":"first","affiliation":[{"name":"Department of Computer Science and Technology, BNRist, Tsinghua University, Beijing, China"}]},{"given":"Lingxiao","family":"Ma","sequence":"additional","affiliation":[{"name":"Microsoft Research, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1039-0386","authenticated-orcid":false,"given":"Ziyan","family":"Fu","sequence":"additional","affiliation":[{"name":"Department of Computer Science and Technology, BNRist, Tsinghua University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-5341-2303","authenticated-orcid":false,"given":"Xiangyu","family":"Li","sequence":"additional","affiliation":[{"name":"Institute for AI Industry Research (AIR), Tsinghua University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1591-2526","authenticated-orcid":false,"given":"Yuanchun","family":"Li","sequence":"additional","affiliation":[{"name":"Institute for AI Industry Research (AIR), Tsinghua University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2782-183X","authenticated-orcid":false,"given":"Ju","family":"Ren","sequence":"additional","affiliation":[{"name":"Department of Computer Science and Technology, BNRist, Tsinghua University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6717-461X","authenticated-orcid":false,"given":"Yaoxue","family":"Zhang","sequence":"additional","affiliation":[{"name":"Department of Computer Science and Technology, BNRist, Tsinghua University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7352-8955","authenticated-orcid":false,"given":"Yunxin","family":"Liu","sequence":"additional","affiliation":[{"name":"Institute for AI Industry Research (AIR), Tsinghua University, Beijing, China"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/TITS.2020.3017505"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1145\/3387514.3405874"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1145\/3581783.3612585"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.23919\/cje.2023.00.288"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-018-0316-z"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1145\/3400713.3400717"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1145\/3307334.3328589"},{"year":"2023","key":"ref8","article-title":"NVIDIA machine learning inference engine"},{"key":"ref9","first-page":"613","article-title":"Clipper: A low-latency online prediction serving system","volume-title":"Proc. 14th USENIX Symp. Networked Syst. Des. Implementation","author":"Crankshaw","year":"2017"},{"year":"2023","key":"ref10","article-title":"Tensorflow serving"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1145\/3600006.3613163"},{"key":"ref12","first-page":"499","article-title":"PipeSwitch: Fast pipelined context switching for deep learning applications","volume-title":"Proc. 14th USENIX Symp. Operating Syst. Des. Implementation","author":"Bai","year":"2020"},{"key":"ref13","first-page":"69","article-title":"Transparent GPU sharing in container clouds for deep learning workloads","volume-title":"Proc. 20th USENIX Symp. Netw. Syst. Des. Implementation","author":"Wu","year":"2023"},{"key":"ref14","first-page":"397","article-title":"INFaaS: Automated model-less inference serving","volume-title":"Proc. 2021 USENIX Annu. Tech. Conf.","author":"Romero","year":"2021"},{"key":"ref15","article-title":"A survey ofmulti-tenant deep learning inference on GPU","volume-title":"Proc. MLSys\u201922 Workshop Cloud Intell.\/AIOps, Co-Located MLSys","author":"Yu","year":"2022"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1145\/3552326.3567499"},{"key":"ref17","first-page":"947","article-title":"Analysis of Large-ScaleMulti-TenantGPU clusters for DNN training workloads","volume-title":"Proc. 2019 USENIX Annu. Tech. Conf.","author":"Jeon","year":"2019"},{"year":"2023","key":"ref18","article-title":"NVIDIA MPS: Multi-process service"},{"year":"2023","key":"ref19","article-title":"NVIDIA MIG: Multi-instance GPU"},{"key":"ref20","first-page":"539","article-title":"Microsecond-scale preemption for concurrent GPU-acceleratedDNN inferences","volume-title":"Proc. 16th USENIX Symp. Operating Syst. Des. Implementation","author":"Han","year":"2022"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1145\/3627703.3629578"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA56546.2023.10071121"},{"year":"2023","key":"ref23","article-title":"NVIDIA cuda basic linear algebra subroutine library"},{"year":"2023","key":"ref24","article-title":"NVIDIA cuda deep neural network library"},{"key":"ref25","first-page":"12156","article-title":"DAGs with no Curl: An efficient DAG structure learning approach","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Yu","year":"2021"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1145\/3453483.3454083"},{"key":"ref27","first-page":"1","article-title":"Apollo: Automatic partition-based operator fusion through layer by layer optimization","volume-title":"Proc. Mach. Learn. Syst.","volume":"4","author":"Zhao","year":"2022"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1145\/3458817.3476143"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1145\/3037697.3037700"},{"key":"ref30","first-page":"663","article-title":"AlpaServe: Statistical multiplexing with model parallelism for deep learning serving","volume-title":"Proc. 17th USENIX Symp. Operating Syst. Des. Implementation","author":"Li","year":"2023"},{"key":"ref31","first-page":"183","article-title":"DVABatch: Diversity-aware Multi-Entry Multi-Exit batching for efficient processing of DNN services on GPUs","volume-title":"Proc. 2022 USENIX Annu. Tech. Conf.","author":"Cui","year":"2022"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1145\/3341301.3359658"},{"key":"ref33","first-page":"787","article-title":"SHEPHERD: Serving DNNs in the wild","volume-title":"Proc. 20th USENIX Symp. Networked Syst. Des. Implementation","author":"Zhang","year":"2023"},{"key":"ref34","first-page":"443","article-title":"Serving DNNs like clockwork: Performance predictability from the bottom up","volume-title":"Proc. 14th USENIX Symp. Operating Syst. Des. Implementation","author":"Gujarati","year":"2020"},{"key":"ref35","first-page":"973","article-title":"Gemel: Model merging for memory-efficient, real-time video analytics at the edge","volume-title":"Proc. 20th USENIX Symp. Netw. Syst. Des. Implementation","author":"Padmanabhan","year":"2023"},{"article-title":"Missile: Fine-grained, hardware-level GPU resource isolation for multi-tenant DNN inference","year":"2024","author":"Zhang","key":"ref36"},{"key":"ref37","first-page":"947","article-title":"USHER: Holistic interference avoidance for resource optimized ML inference","volume-title":"Proc. 18th USENIX Symp. Operating Syst. Des. Implementation","author":"Shubha","year":"2024"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1145\/3627703.3650074"},{"key":"ref39","first-page":"881","article-title":"Rammer: Enabling holistic deep learning compiler optimizations with rTasks","volume-title":"Proc. 14th USENIX Symp. Operating Syst. Des. Implementation","author":"Ma","year":"2020"},{"key":"ref40","first-page":"8343","article-title":"Nimble: Lightweight and parallel GPU task scheduling for deep learning","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Kwon","year":"2020"},{"key":"ref41","first-page":"578","article-title":"TVM: An automated end-to-end optimizing compiler for deep learning","volume-title":"Proc. 13th USENIX Symp. Operating Syst. Des. Implementation","author":"Chen","year":"2018"},{"key":"ref42","first-page":"681","article-title":"Cocktailer: Analyzing and optimizing dynamic control flow in deep learning","volume-title":"Proc. 17th USENIX Symp. Operating Syst. Des. Implementation","author":"Zhang","year":"2023"},{"key":"ref43","first-page":"8026","article-title":"PyTorch: An imperative style, high-performance deep learning library","volume-title":"Proc. 33rd Int. Conf. Neural Inf. Process. Syst.","author":"Paszke","year":"2019"},{"year":"2023","key":"ref44","article-title":"Cuda docs: Develop, optimize and deploy GPU-accelerated apps"},{"key":"ref45","first-page":"533","article-title":"AntMan: Dynamic scaling on GPU clusters for deep learning","volume-title":"Proc. 14th USENIX Symp. Operating Syst. Des. Implementation","author":"Xiao","year":"2020"},{"article-title":"AMD ROCm platform documentation","year":"2023","author":"ROCm","key":"ref46"},{"key":"ref47","first-page":"233","article-title":"ROLLER: Fast and efficient tensor compilation for deep learning","volume-title":"Proc. 16th USENIX Symp. Operating Syst. Des. Implementation","author":"Zhu","year":"2022"}],"container-title":["IEEE Transactions on Mobile Computing"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/7755\/11229982\/11096018.pdf?arnumber=11096018","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,6]],"date-time":"2025-11-06T18:55:18Z","timestamp":1762455318000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11096018\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,12]]},"references-count":47,"journal-issue":{"issue":"12"},"URL":"https:\/\/doi.org\/10.1109\/tmc.2025.3592647","relation":{},"ISSN":["1536-1233","1558-0660","2161-9875"],"issn-type":[{"type":"print","value":"1536-1233"},{"type":"electronic","value":"1558-0660"},{"type":"electronic","value":"2161-9875"}],"subject":[],"published":{"date-parts":[[2025,12]]}}}