{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,24]],"date-time":"2026-01-24T15:16:03Z","timestamp":1769267763718,"version":"3.49.0"},"reference-count":52,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"5","license":[{"start":{"date-parts":[[2021,5,1]],"date-time":"2021-05-01T00:00:00Z","timestamp":1619827200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2021,5,1]],"date-time":"2021-05-01T00:00:00Z","timestamp":1619827200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2021,5,1]],"date-time":"2021-05-01T00:00:00Z","timestamp":1619827200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"Hong Kong RGC Research Impact Fund","award":["R5060-19"],"award-info":[{"award-number":["R5060-19"]}]},{"name":"Hong Kong RGC Research Impact Fund","award":["R5034-18"],"award-info":[{"award-number":["R5034-18"]}]},{"name":"General Research Fund","award":["152221\/19E"],"award-info":[{"award-number":["152221\/19E"]}]},{"name":"Collaborative Research Fund","award":["C5026-18G"],"award-info":[{"award-number":["C5026-18G"]}]},{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["61872310"],"award-info":[{"award-number":["61872310"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100002858","name":"China Postdoctoral Science Foundation","doi-asserted-by":"publisher","award":["2019M661709"],"award-info":[{"award-number":["2019M661709"]}],"id":[{"id":"10.13039\/501100002858","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Shanghai Trusted Industrial Control Platform"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Parallel Distrib. Syst."],"published-print":{"date-parts":[[2021,5,1]]},"DOI":"10.1109\/tpds.2020.3040601","type":"journal-article","created":{"date-parts":[[2020,11,25]],"date-time":"2020-11-25T20:58:14Z","timestamp":1606337894000},"page":"1030-1043","source":"Crossref","is-referenced-by-count":33,"title":["Petrel: Heterogeneity-Aware Distributed Deep Learning Via Hybrid Synchronization"],"prefix":"10.1109","volume":"32","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-2052-638X","authenticated-orcid":false,"given":"Qihua","family":"Zhou","sequence":"first","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9831-2202","authenticated-orcid":false,"given":"Song","family":"Guo","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7538-1985","authenticated-orcid":false,"given":"Zhihao","family":"Qu","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4981-0496","authenticated-orcid":false,"given":"Peng","family":"Li","sequence":"additional","affiliation":[]},{"given":"Li","family":"Li","sequence":"additional","affiliation":[]},{"given":"Minyi","family":"Guo","sequence":"additional","affiliation":[]},{"given":"Kun","family":"Wang","sequence":"additional","affiliation":[]}],"member":"263","reference":[{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/5.726791"},{"key":"ref38","first-page":"1097","article-title":"ImageNet classification with deep convolutional neural networks","author":"krizhevsky","year":"2012","journal-title":"Proc 25th Int Conf Neural Inf Process Syst"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.308"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1145\/2987550.2987554"},{"key":"ref31","first-page":"1","article-title":"Automatic differentiation in PyTorch","author":"paszke","year":"2017","journal-title":"Proc NIPS Autodiff Workshop"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2019.2918951"},{"key":"ref37","year":"0"},{"key":"ref36","article-title":"Omnivore: An optimizer for multi-device deep learning on CPUs and GPUs","author":"hadjis","year":"2016","journal-title":"arXiv 1606 04487"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1145\/3126908.3126916"},{"key":"ref34","article-title":"Learning multiple layers of features from tiny images","author":"krizhevsky","year":"0"},{"key":"ref28","first-page":"521","article-title":"Distance metric learning, with application to clustering with side-information","author":"xing","year":"2002","journal-title":"Proc 15th Int Conf Neural Inf Proc"},{"key":"ref27","first-page":"1928","article-title":"Asynchronous methods for deep reinforcement learning","author":"mnih","year":"2016","journal-title":"Proc 33rd Int Conf Mach Learn"},{"key":"ref29","article-title":"An efficient framework for clustered federated learning","author":"ghosh","year":"2020","journal-title":"arXiv 2006 04088"},{"key":"ref2","first-page":"265","article-title":"TensorFlow: A system for large-scale machine learning","author":"abadi","year":"2016","journal-title":"Proc 11th USENIX Conf Operating Syst Des Implementation"},{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/TBDATA.2015.2472014"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1145\/3079856.3080244"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1145\/1807167.1807184"},{"key":"ref21","first-page":"4243","article-title":"BML: A high-performance, low-cost gradient synchronization algorithm for DML training","author":"wang","year":"2018","journal-title":"Proc 32nd Int Conf Neural Inf Process Syst"},{"key":"ref24","first-page":"693","article-title":"HOGWILD!: A lock-free approach to parallelizing stochastic gradient descent","author":"niu","year":"2011","journal-title":"Proc 24th Int Conf Neural Inf Process Syst"},{"key":"ref23","first-page":"1223","article-title":"Large scale distributed deep networks","author":"dean","year":"2012","journal-title":"Proc 25th Int Conf Neural Inf Process Syst"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1145\/2987550.2987586"},{"key":"ref25","first-page":"2350","article-title":"Staleness-aware async-SGD for distributed deep learning","author":"zhang","year":"2016","journal-title":"Proc 25th Int Joint Conf Artif Intell"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.634"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"ref52","first-page":"1","article-title":"Petrel: Community-aware synchronous parallel for heterogeneous parameter server","author":"zhou","year":"2020","journal-title":"Proc Int Conf Distrib Comput Syst"},{"key":"ref10","first-page":"583","article-title":"Scaling distributed machine learning with the parameter server","author":"li","year":"2014","journal-title":"Proc 11th USENIX Conf Operating Syst Des Implementation"},{"key":"ref11","first-page":"19","article-title":"Communication efficient distributed machine learning with the parameter server","author":"li","year":"2014","journal-title":"Proc 27th Int Conf Neural Inf Process Syst"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1145\/3337821.3337828"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1093\/nsr\/nwx018"},{"key":"ref13","first-page":"561","article-title":"Ray: A distributed framework for emerging AI applications","author":"moritz","year":"2018","journal-title":"Proc 11th USENIX Conf Operating Syst Des Implementation"},{"key":"ref14","article-title":"A berkeley view of systems challenges for AI","author":"stoica","year":"2017","journal-title":"arXiv 1712 05855"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1145\/3267809.3275463"},{"key":"ref16","first-page":"3368","article-title":"Gradient coding: Avoiding stragglers in distributed learning","author":"tandon","year":"2017","journal-title":"Proc 34th Int Conf Mach Learn"},{"key":"ref17","article-title":"Revisiting distributed synchronous SGD","author":"chen","year":"2016","journal-title":"arXiv 1604 00981"},{"key":"ref18","first-page":"181","article-title":"Poseidon: An efficient communication architecture for distributed deep learning on GPU clusters","author":"zhang","year":"2017","journal-title":"Proc USENIX Conf USENIX Annu Tech Conf"},{"key":"ref19","first-page":"595","article-title":"Gandiva: Introspective cluster scheduling for deep learning","author":"xiao","year":"2018","journal-title":"Proc 11th USENIX Conf Operating Syst Des Implementation"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1145\/3187009.3177734"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1145\/3035918.3035933"},{"key":"ref6","first-page":"37","article-title":"Exploiting bounded staleness to speed up big data analytics","author":"cui","year":"2014","journal-title":"Proc USENIX Conf USENIX Annu Tech Conf"},{"key":"ref5","first-page":"629","article-title":"Gaia: Geo-distributed machine learning approaching LAN speeds","author":"hsieh","year":"2017","journal-title":"Proc 10th USENIX Conf Netw Syst Des Implementation"},{"key":"ref8","first-page":"1223","article-title":"More effective distributed ML via a stale synchronous parallel parameter server","author":"ho","year":"2013","journal-title":"Proc 26th Int Conf Neural Inf Process Syst"},{"key":"ref7","first-page":"571","article-title":"Project adam: Building an efficient and scalable deep learning training system","author":"chilimbi","year":"2014","journal-title":"Proc 11th USENIX Conf Operating Syst Des Implementation"},{"key":"ref49","article-title":"Very deep convolutional networks for large-scale image recognition","author":"simonyan","year":"2014","journal-title":"arXiv 1409 1556"},{"key":"ref9","article-title":"MXNet: A flexible and efficient machine learning library for heterogeneous distributed systems","author":"chen","year":"2015","journal-title":"arXiv 1512 01274"},{"key":"ref46","first-page":"281","article-title":"Some methods for classification and analysis of multivariate observations.","author":"macqueen","year":"1967","journal-title":"Proc 5th Berkeley Symp Math Statist Probability"},{"key":"ref45","article-title":"Fashion-MNIST: A novel image dataset for benchmarking machine learning algorithms","author":"xiao","year":"2017","journal-title":"ArXiv 1708 07747"},{"key":"ref48","year":"0"},{"key":"ref47","first-page":"5336","article-title":"Can decentralized algorithms outperform centralized algorithms? A case study for decentralized parallel stochastic gradient descent","author":"lian","year":"2017","journal-title":"Proc 31st Int Conf Neural Inf Process Syst"},{"key":"ref42","volume":"1","author":"oliphant","year":"2006","journal-title":"A Guide to NumPy"},{"key":"ref41","first-page":"2825","article-title":"Scikit-learn: Machine learning in Python","volume":"12","author":"pedregosa","year":"2011","journal-title":"J Mach Learn Res"},{"key":"ref44","first-page":"226","article-title":"A density-based algorithm for discovering clusters in large spatial databases with noise","author":"ester","year":"1996","journal-title":"Proc Int Conf Knowl Disc Data Mining"},{"key":"ref43","year":"0"}],"container-title":["IEEE Transactions on Parallel and Distributed Systems"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/71\/9275496\/09271915.pdf?arnumber=9271915","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2022,5,10]],"date-time":"2022-05-10T14:50:27Z","timestamp":1652194227000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/9271915\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,5,1]]},"references-count":52,"journal-issue":{"issue":"5"},"URL":"https:\/\/doi.org\/10.1109\/tpds.2020.3040601","relation":{},"ISSN":["1045-9219","1558-2183","2161-9883"],"issn-type":[{"value":"1045-9219","type":"print"},{"value":"1558-2183","type":"electronic"},{"value":"2161-9883","type":"electronic"}],"subject":[],"published":{"date-parts":[[2021,5,1]]}}}