{"id":"https://openalex.org/W7161122523","doi":"https://doi.org/10.1145/3746467.3801522","title":"GPU Starvation from the Control Plane: A Systems Study of Reinforcement Learning in Deep Learning Pipelines","display_name":"GPU Starvation from the Control Plane: A Systems Study of Reinforcement Learning in Deep Learning Pipelines","publication_year":2026,"publication_date":"2026-04-23","ids":{"openalex":"https://openalex.org/W7161122523","doi":"https://doi.org/10.1145/3746467.3801522"},"language":null,"primary_location":{"id":"doi:10.1145/3746467.3801522","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3746467.3801522","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 ACM Southeast Conference","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3746467.3801522","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5129705383","display_name":"Michael Seavers","orcid":null},"institutions":[{"id":"https://openalex.org/I94339441","display_name":"Western Kentucky University","ror":"https://ror.org/0446vnd56","country_code":"US","type":"education","lineage":["https://openalex.org/I94339441"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Michael Seavers","raw_affiliation_strings":["Western Kentucky University, Bowling Green, KY, USA"],"raw_orcid":"https://orcid.org/0009-0001-7019-1979","affiliations":[{"raw_affiliation_string":"Western Kentucky University, Bowling Green, KY, USA","institution_ids":["https://openalex.org/I94339441"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034599517","display_name":"Yongyi Gong","orcid":"https://orcid.org/0000-0002-8559-1801"},"institutions":[{"id":"https://openalex.org/I186272606","display_name":"Guangdong University of Foreign Studies","ror":"https://ror.org/00fhc9y79","country_code":"CN","type":"education","lineage":["https://openalex.org/I186272606"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yongyi Gong","raw_affiliation_strings":["Guangdong University of Foreign Studies, Guangzhou, China"],"raw_orcid":"https://orcid.org/0000-0002-8559-1801","affiliations":[{"raw_affiliation_string":"Guangdong University of Foreign Studies, Guangzhou, China","institution_ids":["https://openalex.org/I186272606"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5136110965","display_name":"Qi Li","orcid":"https://orcid.org/0000-0001-5554-5647"},"institutions":[{"id":"https://openalex.org/I94339441","display_name":"Western Kentucky University","ror":"https://ror.org/0446vnd56","country_code":"US","type":"education","lineage":["https://openalex.org/I94339441"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Qi Li","raw_affiliation_strings":["School of Engineering and Applied Sciences, Western Kentucky University, Bowling Green, USA"],"raw_orcid":"https://orcid.org/0000-0001-5554-5647","affiliations":[{"raw_affiliation_string":"School of Engineering and Applied Sciences, Western Kentucky University, Bowling Green, USA","institution_ids":["https://openalex.org/I94339441"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5129705383"],"corresponding_institution_ids":["https://openalex.org/I94339441"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.95016546,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"205","last_page":"210"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.15690000355243683,"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"}},"topics":[{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.15690000355243683,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.1477999985218048,"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/T10101","display_name":"Cloud Computing and Resource Management","score":0.11640000343322754,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/reinforcement-learning","display_name":"Reinforcement learning","score":0.765999972820282},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.6380000114440918},{"id":"https://openalex.org/keywords/starvation","display_name":"Starvation","score":0.5386000275611877},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.5066999793052673},{"id":"https://openalex.org/keywords/pipeline-transport","display_name":"Pipeline transport","score":0.4408999979496002},{"id":"https://openalex.org/keywords/resource","display_name":"Resource (disambiguation)","score":0.427700012922287},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.41830000281333923},{"id":"https://openalex.org/keywords/graphics-processing-unit","display_name":"Graphics processing unit","score":0.39579999446868896}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.765999972820282},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7304999828338623},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.6380000114440918},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5781999826431274},{"id":"https://openalex.org/C2780606744","wikidata":"https://www.wikidata.org/wiki/Q853930","display_name":"Starvation","level":2,"score":0.5386000275611877},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.5066999793052673},{"id":"https://openalex.org/C175309249","wikidata":"https://www.wikidata.org/wiki/Q725864","display_name":"Pipeline transport","level":2,"score":0.4408999979496002},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.427700012922287},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.41830000281333923},{"id":"https://openalex.org/C2779851693","wikidata":"https://www.wikidata.org/wiki/Q183484","display_name":"Graphics processing unit","level":2,"score":0.39579999446868896},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.3874000012874603},{"id":"https://openalex.org/C21442007","wikidata":"https://www.wikidata.org/wiki/Q1027879","display_name":"Graphics","level":2,"score":0.3587999939918518},{"id":"https://openalex.org/C34413123","wikidata":"https://www.wikidata.org/wiki/Q170978","display_name":"Robotics","level":3,"score":0.3467000126838684},{"id":"https://openalex.org/C172658912","wikidata":"https://www.wikidata.org/wiki/Q661613","display_name":"Batch processing","level":2,"score":0.34369999170303345},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.33739998936653137},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2847999930381775},{"id":"https://openalex.org/C107464732","wikidata":"https://www.wikidata.org/wiki/Q235781","display_name":"Adaptive control","level":3,"score":0.2782999873161316},{"id":"https://openalex.org/C138827492","wikidata":"https://www.wikidata.org/wiki/Q6661985","display_name":"Data processing","level":2,"score":0.275299996137619},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.2590999901294708},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.250900000333786}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3746467.3801522","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3746467.3801522","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 ACM Southeast Conference","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3746467.3801522","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3746467.3801522","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 ACM Southeast Conference","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/8","score":0.4746297299861908,"display_name":"Decent work and economic growth"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":6,"referenced_works":["https://openalex.org/W1864199185","https://openalex.org/W2112420033","https://openalex.org/W2886675192","https://openalex.org/W4242175757","https://openalex.org/W4394745249","https://openalex.org/W4414427843"],"related_works":[],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1,12],"(RL)":[2],"is":[3],"increasingly":[4],"deployed":[5],"in":[6,123,172],"the":[7,39,77,163],"control":[8,168],"plane":[9,88],"of":[10,63,76,130],"deep":[11],"systems":[13],"to":[14],"automate":[15],"decisions":[16],"such":[17],"as":[18,85],"batch":[19],"sizing,":[20],"worker":[21],"allocation,":[22],"and":[23,45,107,114,133,159,169],"resource":[24],"scheduling.":[25],"While":[26],"prior":[27],"work":[28],"has":[29],"emphasized":[30],"performance":[31,74,171],"improvements":[32],"enabled":[33],"by":[34,67],"RL,":[35],"its":[36],"impact":[37],"on":[38],"underlying":[40],"Central":[41],"Processing":[42,47],"Unit":[43,48],"(CPU)":[44],"Graphics":[46],"(GPU)":[49],"execution":[50,102],"pipeline":[51,80,170],"remains":[52],"insufficiently":[53],"understood.":[54],"In":[55],"this":[56,124],"paper,":[57],"we":[58,126],"present":[59],"a":[60,73,86,128],"systems-level":[61],"study":[62],"GPU":[64,83],"starvation":[65,84,120,131],"induced":[66],"control-plane":[68],"reinforcement":[69],"learning.":[70],"We":[71],"develop":[72],"model":[75,100],"CPU-GPU":[78],"data":[79],"that":[81,147],"characterizes":[82],"control-data":[87],"imbalance,":[89],"arising":[90],"when":[91],"CPU-side":[92],"production":[93],"time":[94,103],"exceeds":[95],"GPU-side":[96],"service":[97],"time.":[98],"The":[99],"decomposes":[101],"into":[104],"batch-dependent":[105],"computation":[106],"per-iteration":[108],"overhead,":[109],"revealing":[110],"how":[111],"fixed":[112],"overheads":[113],"batch-size":[115,142],"dynamics":[116],"can":[117,152],"nonlinearly":[118],"amplify":[119],"effects.":[121],"Grounded":[122],"model,":[125],"introduce":[127],"taxonomy":[129],"behaviors":[132],"validate":[134],"our":[135],"analysis":[136],"through":[137],"extensive":[138],"experiments":[139],"with":[140],"RL-driven":[141],"control.":[143],"Our":[144],"results":[145],"demonstrate":[146],"even":[148],"lightweight":[149],"RL":[150],"methods":[151],"induce":[153],"complex":[154],"utilization":[155],"patterns,":[156],"including":[157],"persistent":[158],"periodic":[160],"starvation,":[161],"highlighting":[162],"subtle":[164],"interaction":[165],"between":[166],"adaptive":[167],"modern":[173],"AI":[174],"systems.":[175]},"counts_by_year":[],"updated_date":"2026-05-15T06:12:33.780692","created_date":"2026-05-15T00:00:00"}