{"id":"https://openalex.org/W4404386172","doi":"https://doi.org/10.1145/3698038.3698532","title":"Kale: Elastic GPU Scheduling for Online DL Model Training","display_name":"Kale: Elastic GPU Scheduling for Online DL Model Training","publication_year":2024,"publication_date":"2024-11-14","ids":{"openalex":"https://openalex.org/W4404386172","doi":"https://doi.org/10.1145/3698038.3698532"},"language":"en","primary_location":{"id":"doi:10.1145/3698038.3698532","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3698038.3698532","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 Symposium on Cloud Computing","raw_type":"proceedings-article"},"type":"article","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":null,"display_name":"Ziyang Liu","orcid":"https://orcid.org/0009-0005-2199-6374"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Ziyang Liu","raw_affiliation_strings":["Beihang University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5050796169","display_name":"Renyu Yang","orcid":"https://orcid.org/0000-0001-6334-4925"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Renyu Yang","raw_affiliation_strings":["Beihang University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Jin Ouyang","orcid":"https://orcid.org/0009-0003-6591-7717"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jin Ouyang","raw_affiliation_strings":["Kuaishou Inc., Beijing, China"],"affiliations":[{"raw_affiliation_string":"Kuaishou Inc., Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Weihan Jiang","orcid":"https://orcid.org/0009-0001-2101-1302"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Weihan Jiang","raw_affiliation_strings":["Beihang University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021414951","display_name":"T Ye","orcid":null},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tianyu Ye","raw_affiliation_strings":["Beihang University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031176052","display_name":"Menghao Zhang","orcid":"https://orcid.org/0000-0001-5274-5512"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Menghao Zhang","raw_affiliation_strings":["Beihang University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102704686","display_name":"Sui Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sui Huang","raw_affiliation_strings":["Kuaishou Inc., Beijing, China"],"affiliations":[{"raw_affiliation_string":"Kuaishou Inc., Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Jiaming Huang","orcid":"https://orcid.org/0009-0008-2631-4979"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiaming Huang","raw_affiliation_strings":["Kuaishou Inc., Beijing, China"],"affiliations":[{"raw_affiliation_string":"Kuaishou Inc., Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070531404","display_name":"Chengru Song","orcid":"https://orcid.org/0009-0009-3826-8436"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chengru Song","raw_affiliation_strings":["Kuaishou Inc., Beijing, China"],"affiliations":[{"raw_affiliation_string":"Kuaishou Inc., Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102941781","display_name":"Di Zhang","orcid":"https://orcid.org/0009-0006-5475-2728"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Di Zhang","raw_affiliation_strings":["Kuaishou Inc., Beijing, China"],"affiliations":[{"raw_affiliation_string":"Kuaishou Inc., Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066123431","display_name":"Tianyu Wo","orcid":"https://orcid.org/0000-0002-5331-3364"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tianyu Wo","raw_affiliation_strings":["Beihang University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5028857237","display_name":"Chunming Hu","orcid":"https://orcid.org/0000-0003-3473-9703"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chunming Hu","raw_affiliation_strings":["Beihang University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":12,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I82880672"],"apc_list":null,"apc_paid":null,"fwci":1.3601,"has_fulltext":false,"cited_by_count":4,"citation_normalized_percentile":{"value":0.83105416,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":97,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"36","last_page":"51"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10715","display_name":"Distributed and Parallel Computing Systems","score":0.9976000189781189,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T10715","display_name":"Distributed and Parallel Computing Systems","score":0.9976000189781189,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9768000245094299,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10052","display_name":"Medical Image Segmentation Techniques","score":0.9765999913215637,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7860708832740784},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.5906251668930054},{"id":"https://openalex.org/keywords/general-purpose-computing-on-graphics-processing-units","display_name":"General-purpose computing on graphics processing units","score":0.4958711564540863},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.4709183871746063},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.4660201072692871},{"id":"https://openalex.org/keywords/computer-graphics","display_name":"Computer graphics (images)","score":0.38036486506462097},{"id":"https://openalex.org/keywords/mathematical-optimization","display_name":"Mathematical optimization","score":0.10691729187965393},{"id":"https://openalex.org/keywords/graphics","display_name":"Graphics","score":0.07112371921539307}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7860708832740784},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.5906251668930054},{"id":"https://openalex.org/C50630238","wikidata":"https://www.wikidata.org/wiki/Q971505","display_name":"General-purpose computing on graphics processing units","level":3,"score":0.4958711564540863},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4709183871746063},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.4660201072692871},{"id":"https://openalex.org/C121684516","wikidata":"https://www.wikidata.org/wiki/Q7600677","display_name":"Computer graphics (images)","level":1,"score":0.38036486506462097},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.10691729187965393},{"id":"https://openalex.org/C21442007","wikidata":"https://www.wikidata.org/wiki/Q1027879","display_name":"Graphics","level":2,"score":0.07112371921539307},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.0},{"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:10.1145/3698038.3698532","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3698038.3698532","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 Symposium on Cloud Computing","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.4399999976158142,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":35,"referenced_works":["https://openalex.org/W95608104","https://openalex.org/W2132782512","https://openalex.org/W2512971201","https://openalex.org/W2723293840","https://openalex.org/W2798515322","https://openalex.org/W2899824923","https://openalex.org/W2973172293","https://openalex.org/W2994850640","https://openalex.org/W3023238978","https://openalex.org/W3027366302","https://openalex.org/W3036320503","https://openalex.org/W3043004181","https://openalex.org/W3080510735","https://openalex.org/W3098468692","https://openalex.org/W3121689374","https://openalex.org/W3130934537","https://openalex.org/W3134991928","https://openalex.org/W3155949183","https://openalex.org/W3159243351","https://openalex.org/W3177318507","https://openalex.org/W3208695543","https://openalex.org/W3216508375","https://openalex.org/W4205916020","https://openalex.org/W4213132668","https://openalex.org/W4290945834","https://openalex.org/W4290991121","https://openalex.org/W4293731856","https://openalex.org/W4308426177","https://openalex.org/W4308426645","https://openalex.org/W4316252403","https://openalex.org/W4386497355","https://openalex.org/W4390263770","https://openalex.org/W4399695571","https://openalex.org/W4401857054","https://openalex.org/W4401863356"],"related_works":["https://openalex.org/W2505380084","https://openalex.org/W230091440","https://openalex.org/W4400333498","https://openalex.org/W2233261550","https://openalex.org/W2810751659","https://openalex.org/W258997015","https://openalex.org/W2997094352","https://openalex.org/W2086739451","https://openalex.org/W3183233360","https://openalex.org/W1980160788"],"abstract_inverted_index":{"Large-scale":[0],"GPU":[1,80,184],"clusters":[2,185],"have":[3],"been":[4,179],"widely":[5],"used":[6],"for":[7,29,125],"effectively":[8],"training":[9,32,131],"both":[10],"online":[11,57,88],"and":[12,95,159,163,186,192],"offline":[13,30],"deep":[14],"learning":[15],"(DL)":[16],"jobs.":[17],"However,":[18],"elastic":[19,79],"scheduling":[20,81],"in":[21,56],"most":[22],"cases":[23],"of":[24,39,49,87,103],"resource":[25,34,64],"schedulers":[26],"is":[27,36,45,122],"dedicated":[28],"model":[31,58,69,90],"where":[33],"adjustment":[35],"planned":[37],"ahead":[38],"time.":[40],"The":[41],"native":[42],"autoscaling":[43,152],"policy":[44],"on":[46],"the":[47,85,101,109,135,144,149,156,168],"basis":[48],"pre-defined":[50],"threshold":[51],"and,":[52],"if":[53],"applied":[54],"directly":[55],"training,":[59],"often":[60],"suffers":[61],"from":[62,172],"belated":[63],"adjustment,":[65],"leading":[66],"to":[67,83,174],"diminished":[68],"accuracy.":[70],"In":[71],"this":[72],"paper,":[73],"we":[74],"present":[75],"Kale,":[76],"a":[77],"novel":[78],"system":[82],"improve":[84],"performance":[86],"DL":[89],"training.":[91],"Through":[92],"traffic":[93],"forecasting":[94],"resource-throughput":[96],"modeling,":[97],"Kale":[98,141,154,177],"automatically":[99],"pinpoints":[100],"number":[102],"required":[104],"GPUs":[105],"that":[106,140],"best":[107],"accommodate":[108],"on-the-fly":[110],"data":[111,119],"samples":[112,128],"before":[113],"performing":[114],"stabilized":[115],"autoscaling.":[116],"An":[117],"advanced":[118],"shuffling":[120],"strategy":[121],"further":[123],"employed":[124],"balancing":[126],"uneven":[127],"among":[129],"different":[130],"workers,":[132],"thereby":[133],"improving":[134],"runtime":[136],"efficacy.":[137],"Experiments":[138],"show":[139],"substantially":[142],"outperforms":[143],"state-of-the-art":[145],"solutions.":[146],"Compared":[147],"with":[148],"default":[150],"HPA":[151],"strategy,":[153],"reduces":[155],"accumulated":[157],"lag":[158],"downtime":[160],"by":[161],"69.2%":[162],"33.1%,":[164],"respectively,":[165],"whilst":[166],"lowering":[167],"SLO":[169],"violation":[170],"rate":[171],"19.57%":[173],"just":[175],"2.6%.":[176],"has":[178],"deployed":[180],"at":[181,194],"Kuaishou's":[182],"production-level":[183],"successfully":[187],"underpins":[188],"real-time":[189],"video":[190],"recommendation":[191],"advertisement":[193],"scale.":[195]},"counts_by_year":[{"year":2025,"cited_by_count":4}],"updated_date":"2026-04-17T18:11:37.981687","created_date":"2025-10-10T00:00:00"}