{"id":"https://openalex.org/W4401176258","doi":"https://doi.org/10.1145/3651890.3672228","title":"Accelerating Model Training in Multi-cluster Environments with Consumer-grade GPUs","display_name":"Accelerating Model Training in Multi-cluster Environments with Consumer-grade GPUs","publication_year":2024,"publication_date":"2024-07-31","ids":{"openalex":"https://openalex.org/W4401176258","doi":"https://doi.org/10.1145/3651890.3672228"},"language":"en","primary_location":{"id":"doi:10.1145/3651890.3672228","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3651890.3672228","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM SIGCOMM 2024 Conference","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5020511196","display_name":"Hwijoon Lim","orcid":"https://orcid.org/0000-0002-9872-6234"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Hwijoon Lim","raw_affiliation_strings":["KAIST, Daejeon, Korea, South ? Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0002-9872-6234","affiliations":[{"raw_affiliation_string":"KAIST, Daejeon, Korea, South ? Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029128283","display_name":"Juncheol Ye","orcid":"https://orcid.org/0009-0008-7673-1279"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Juncheol Ye","raw_affiliation_strings":["KAIST, Daejeon, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0008-7673-1279","affiliations":[{"raw_affiliation_string":"KAIST, Daejeon, Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5058967466","display_name":"Sangeetha Abdu Jyothi","orcid":"https://orcid.org/0009-0000-0503-4478"},"institutions":[{"id":"https://openalex.org/I204250578","display_name":"University of California, Irvine","ror":"https://ror.org/04gyf1771","country_code":"US","type":"education","lineage":["https://openalex.org/I204250578"]},{"id":"https://openalex.org/I74786994","display_name":"Kitware (United States)","ror":"https://ror.org/02s2acn37","country_code":"US","type":"company","lineage":["https://openalex.org/I74786994"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Sangeetha Abdu Jyothi","raw_affiliation_strings":["UC Irvine, VMware Research, Irvine, California, United States of America"],"raw_orcid":"https://orcid.org/0009-0000-0503-4478","affiliations":[{"raw_affiliation_string":"UC Irvine, VMware Research, Irvine, California, United States of America","institution_ids":["https://openalex.org/I204250578","https://openalex.org/I74786994"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5015247369","display_name":"Dongsu Han","orcid":"https://orcid.org/0000-0001-6922-7244"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Dongsu Han","raw_affiliation_strings":["KAIST, Daejeon, Korea, South ? Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0001-6922-7244","affiliations":[{"raw_affiliation_string":"KAIST, Daejeon, Korea, South ? Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":16,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"707","last_page":"720"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9954000115394592,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9954000115394592,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9948999881744385,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9937999844551086,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7564435005187988},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.6567440629005432},{"id":"https://openalex.org/keywords/cluster","display_name":"Cluster (spacecraft)","score":0.5692634582519531},{"id":"https://openalex.org/keywords/gpu-cluster","display_name":"GPU cluster","score":0.5488657355308533},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.4750387966632843},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.40092551708221436},{"id":"https://openalex.org/keywords/computer-graphics","display_name":"Computer graphics (images)","score":0.3635094165802002},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.2368377447128296},{"id":"https://openalex.org/keywords/cuda","display_name":"CUDA","score":0.19787809252738953}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7564435005187988},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.6567440629005432},{"id":"https://openalex.org/C164866538","wikidata":"https://www.wikidata.org/wiki/Q367351","display_name":"Cluster (spacecraft)","level":2,"score":0.5692634582519531},{"id":"https://openalex.org/C2781335571","wikidata":"https://www.wikidata.org/wiki/Q2633544","display_name":"GPU cluster","level":3,"score":0.5488657355308533},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4750387966632843},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.40092551708221436},{"id":"https://openalex.org/C121684516","wikidata":"https://www.wikidata.org/wiki/Q7600677","display_name":"Computer graphics (images)","level":1,"score":0.3635094165802002},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.2368377447128296},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.19787809252738953},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3651890.3672228","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3651890.3672228","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM SIGCOMM 2024 Conference","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3419938175","display_name":null,"funder_award_id":"RS-2024-00340099","funder_id":"https://openalex.org/F4320322120","funder_display_name":"National Research Foundation of Korea"}],"funders":[{"id":"https://openalex.org/F4320322120","display_name":"National Research Foundation of Korea","ror":"https://ror.org/013aysd81"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":16,"referenced_works":["https://openalex.org/W2060393849","https://openalex.org/W2087402357","https://openalex.org/W2108598243","https://openalex.org/W2194775991","https://openalex.org/W2407022425","https://openalex.org/W2734941459","https://openalex.org/W2975712713","https://openalex.org/W2982475424","https://openalex.org/W2997709794","https://openalex.org/W3046364876","https://openalex.org/W3091097978","https://openalex.org/W3138516171","https://openalex.org/W3156794839","https://openalex.org/W3202009823","https://openalex.org/W4372267133","https://openalex.org/W4389109114"],"related_works":["https://openalex.org/W2082176405","https://openalex.org/W2238531340","https://openalex.org/W2067902980","https://openalex.org/W2017587301","https://openalex.org/W1827514603","https://openalex.org/W2965037703","https://openalex.org/W2972376311","https://openalex.org/W1904802460","https://openalex.org/W3215286539","https://openalex.org/W2086392083"],"abstract_inverted_index":{"Rapid":[0],"advances":[1],"in":[2,47,72,86,141,200],"machine":[3],"learning":[4],"necessitate":[5],"significant":[6],"computing":[7],"power":[8],"and":[9,32,61,99,108,171,181,192],"memory":[10],"for":[11],"training,":[12],"which":[13],"is":[14],"accessible":[15],"only":[16,26],"to":[17,38,58,83,104,123,135,148,169,195],"large":[18],"corporations":[19],"today.":[20],"Small-scale":[21],"players":[22],"like":[23],"academics":[24],"often":[25],"have":[27],"consumer-grade":[28],"GPU":[29,36,106],"clusters":[30,194],"locally":[31],"can":[33,157],"afford":[34],"cloud":[35,193],"instances":[37],"a":[39,64,78,100,203],"limited":[40],"extent.":[41],"However,":[42],"training":[43,60,70,111,121,127,146,205],"performance":[44],"significantly":[45],"degrades":[46],"this":[48,52],"multi-cluster":[49],"setting.":[50],"In":[51],"paper,":[53],"we":[54,156],"identify":[55],"unique":[56],"opportunities":[57],"accelerate":[59],"propose":[62],"StellaTrain,":[63],"holistic":[65],"framework":[66],"that":[67,131,155],"achieves":[68,133],"near-optimal":[69],"speeds":[71],"multi-cloud":[73],"environments.":[74],"StellaTrain":[75,89,117,132,153,186],"dynamically":[76],"adapts":[77],"combination":[79],"of":[80,178,207],"acceleration":[81,92],"techniques":[82,93],"minimize":[84],"time-to-accuracy":[85],"model":[87],"training.":[88],"introduces":[90],"novel":[91],"such":[94],"as":[95],"cache-aware":[96],"gradient":[97],"compression":[98],"CPU-based":[101],"sparse":[102],"optimizer":[103],"maximize":[105],"utilization":[107],"optimize":[109,124],"the":[110,114,120,125,160,175],"pipeline.":[112],"With":[113],"optimized":[115],"pipeline,":[116],"holistically":[118],"determines":[119],"configurations":[122,147],"total":[126],"time.":[128],"We":[129],"show":[130],"up":[134,168],"104\u00d7":[136],"speedup":[137],"over":[138],"PyTorch":[139],"DDP":[140],"inter-cluster":[142],"settings":[143],"by":[144,198],"adapting":[145],"fluctuating":[149],"dynamic":[150],"network":[151,162,176],"bandwidth.":[152],"demonstrates":[154],"cope":[158],"with":[159,202],"scarce":[161],"bandwidth":[163],"through":[164],"systematic":[165],"optimization,":[166],"achieving":[167],"257.3\u00d7":[170],"78.1\u00d7":[172],"speed-ups":[173],"on":[174],"bandwidths":[177],"100":[179],"Mbps":[180],"500":[182],"Mbps,":[183],"respectively.":[184],"Finally,":[185],"enables":[187],"efficient":[188],"co-training":[189],"using":[190],"on-premises":[191],"reduce":[196],"costs":[197],"64.5%":[199],"conjunction":[201],"reduced":[204],"time":[206],"28.9%.":[208]},"counts_by_year":[{"year":2026,"cited_by_count":6},{"year":2025,"cited_by_count":10}],"updated_date":"2026-07-22T07:51:19.307946","created_date":"2025-10-10T00:00:00"}