{"id":"https://openalex.org/W4413360549","doi":"https://doi.org/10.1109/cloud67622.2025.00036","title":"Mind the Memory Gap: Unveiling GPU Bottlenecks in Large-Batch LLM Inference","display_name":"Mind the Memory Gap: Unveiling GPU Bottlenecks in Large-Batch LLM Inference","publication_year":2025,"publication_date":"2025-07-07","ids":{"openalex":"https://openalex.org/W4413360549","doi":"https://doi.org/10.1109/cloud67622.2025.00036"},"language":"en","primary_location":{"id":"doi:10.1109/cloud67622.2025.00036","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cloud67622.2025.00036","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 18th International Conference on Cloud Computing (CLOUD)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://upcommons.upc.edu/bitstreams/82e2be60-b600-4fa1-90ff-08d66f1cac7a/download","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5026630988","display_name":"Pol G. Recasens","orcid":"https://orcid.org/0009-0008-5535-0012"},"institutions":[{"id":"https://openalex.org/I2799803557","display_name":"Barcelona Supercomputing Center","ror":"https://ror.org/05sd8tv96","country_code":"ES","type":"facility","lineage":["https://openalex.org/I2799803557","https://openalex.org/I9617848"]},{"id":"https://openalex.org/I9617848","display_name":"Universitat Polit\u00e8cnica de Catalunya","ror":"https://ror.org/03mb6wj31","country_code":"ES","type":"education","lineage":["https://openalex.org/I9617848"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Pol G. Recasens","raw_affiliation_strings":["Barcelona Supercomputing Center (BSC)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Barcelona Supercomputing Center (BSC)","institution_ids":["https://openalex.org/I2799803557","https://openalex.org/I9617848"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Ferran Agullo","orcid":null},"institutions":[{"id":"https://openalex.org/I2799803557","display_name":"Barcelona Supercomputing Center","ror":"https://ror.org/05sd8tv96","country_code":"ES","type":"facility","lineage":["https://openalex.org/I2799803557","https://openalex.org/I9617848"]},{"id":"https://openalex.org/I9617848","display_name":"Universitat Polit\u00e8cnica de Catalunya","ror":"https://ror.org/03mb6wj31","country_code":"ES","type":"education","lineage":["https://openalex.org/I9617848"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Ferran Agullo","raw_affiliation_strings":["Barcelona Supercomputing Center (BSC)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Barcelona Supercomputing Center (BSC)","institution_ids":["https://openalex.org/I2799803557","https://openalex.org/I9617848"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5004237513","display_name":"Yue Zhu","orcid":"https://orcid.org/0009-0005-2207-1904"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yue Zhu","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100337659","display_name":"Chen Wang","orcid":"https://orcid.org/0000-0003-1963-4954"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen Wang","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023638218","display_name":"Eun Kyung Lee","orcid":"https://orcid.org/0000-0002-2335-4798"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Eun Kyung Lee","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5086114919","display_name":"Olivier Tardieu","orcid":"https://orcid.org/0000-0002-8377-6757"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Olivier Tardieu","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039375454","display_name":"Jordi Torres","orcid":"https://orcid.org/0000-0003-1963-7418"},"institutions":[{"id":"https://openalex.org/I2799803557","display_name":"Barcelona Supercomputing Center","ror":"https://ror.org/05sd8tv96","country_code":"ES","type":"facility","lineage":["https://openalex.org/I2799803557","https://openalex.org/I9617848"]},{"id":"https://openalex.org/I9617848","display_name":"Universitat Polit\u00e8cnica de Catalunya","ror":"https://ror.org/03mb6wj31","country_code":"ES","type":"education","lineage":["https://openalex.org/I9617848"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Jordi Torres","raw_affiliation_strings":["Barcelona Supercomputing Center (BSC)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Barcelona Supercomputing Center (BSC)","institution_ids":["https://openalex.org/I2799803557","https://openalex.org/I9617848"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5074281010","display_name":"Josep Ll. Berral","orcid":"https://orcid.org/0000-0003-3037-3580"},"institutions":[{"id":"https://openalex.org/I2799803557","display_name":"Barcelona Supercomputing Center","ror":"https://ror.org/05sd8tv96","country_code":"ES","type":"facility","lineage":["https://openalex.org/I2799803557","https://openalex.org/I9617848"]},{"id":"https://openalex.org/I9617848","display_name":"Universitat Polit\u00e8cnica de Catalunya","ror":"https://ror.org/03mb6wj31","country_code":"ES","type":"education","lineage":["https://openalex.org/I9617848"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Josep Ll. Berral","raw_affiliation_strings":["Barcelona Supercomputing Center (BSC)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Barcelona Supercomputing Center (BSC)","institution_ids":["https://openalex.org/I2799803557","https://openalex.org/I9617848"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":19.9092,"has_fulltext":true,"cited_by_count":13,"citation_normalized_percentile":{"value":0.99543813,"is_in_top_1_percent":true,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":96,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"277","last_page":"287"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12535","display_name":"Machine Learning and Data Classification","score":0.8371999859809875,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.8371999859809875,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.7968999743461609,"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/T10862","display_name":"AI in cancer detection","score":0.7516999840736389,"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":0.6957837343215942},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6031365394592285},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5449652671813965},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3622744679450989}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6957837343215942},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6031365394592285},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5449652671813965},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3622744679450989}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/cloud67622.2025.00036","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cloud67622.2025.00036","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 18th International Conference on Cloud Computing (CLOUD)","raw_type":"proceedings-article"},{"id":"pmh:oai:upcommons.upc.edu:2117/441242","is_oa":true,"landing_page_url":"https://hdl.handle.net/2117/441242","pdf_url":"https://upcommons.upc.edu/bitstreams/82e2be60-b600-4fa1-90ff-08d66f1cac7a/download","source":{"id":"https://openalex.org/S4377196262","display_name":"UPCommons institutional repository (Universitat Polit\u00e8cnica de Catalunya)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I9617848","host_organization_name":"Universitat Polit\u00e8cnica de Catalunya","host_organization_lineage":["https://openalex.org/I9617848"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"info:eu-repo/semantics/conferenceObject"}],"best_oa_location":{"id":"pmh:oai:upcommons.upc.edu:2117/441242","is_oa":true,"landing_page_url":"https://hdl.handle.net/2117/441242","pdf_url":"https://upcommons.upc.edu/bitstreams/82e2be60-b600-4fa1-90ff-08d66f1cac7a/download","source":{"id":"https://openalex.org/S4377196262","display_name":"UPCommons institutional repository (Universitat Polit\u00e8cnica de Catalunya)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I9617848","host_organization_name":"Universitat Polit\u00e8cnica de Catalunya","host_organization_lineage":["https://openalex.org/I9617848"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"info:eu-repo/semantics/conferenceObject"},"sustainable_development_goals":[{"score":0.5699999928474426,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[{"id":"https://openalex.org/G121809054","display_name":null,"funder_award_id":"MCIN/AEI/10","funder_id":"https://openalex.org/F4320322930","funder_display_name":"Ministerio de Ciencia e Innovaci\u00f3n"},{"id":"https://openalex.org/G1321411063","display_name":null,"funder_award_id":"2021-SGR-00478","funder_id":"https://openalex.org/F4320334830","funder_display_name":"Ag\u00e8ncia de Gesti\u00f3 d'Ajuts Universitaris i de Recerca"},{"id":"https://openalex.org/G1622782223","display_name":null,"funder_award_id":"MCIN/AEI/10","funder_id":"https://openalex.org/F4320335322","funder_display_name":"European Regional Development Fund"},{"id":"https://openalex.org/G1855990336","display_name":null,"funder_award_id":".13039/501100011033","funder_id":"https://openalex.org/F4320322930","funder_display_name":"Ministerio de Ciencia e Innovaci\u00f3n"},{"id":"https://openalex.org/G2262748287","display_name":null,"funder_award_id":"501100011033","funder_id":"https://openalex.org/F4320335598","funder_display_name":"Agencia Estatal de Investigaci\u00f3n"},{"id":"https://openalex.org/G2660406951","display_name":null,"funder_award_id":"/ AEI/10","funder_id":"https://openalex.org/F4320322930","funder_display_name":"Ministerio de Ciencia e Innovaci\u00f3n"},{"id":"https://openalex.org/G2778551575","display_name":null,"funder_award_id":"2021-SGR-00478","funder_id":"https://openalex.org/F4320321505","funder_display_name":"Generalitat de Catalunya"},{"id":"https://openalex.org/G3005324705","display_name":null,"funder_award_id":"501100011033","funder_id":"https://openalex.org/F4320321505","funder_display_name":"Generalitat de Catalunya"},{"id":"https://openalex.org/G3480869486","display_name":null,"funder_award_id":"13039","funder_id":"https://openalex.org/F4320335598","funder_display_name":"Agencia Estatal de Investigaci\u00f3n"},{"id":"https://openalex.org/G425668148","display_name":null,"funder_award_id":".13039/501100011033","funder_id":"https://openalex.org/F4320335598","funder_display_name":"Agencia Estatal de Investigaci\u00f3n"},{"id":"https://openalex.org/G451917667","display_name":null,"funder_award_id":"13039/501100011033","funder_id":"https://openalex.org/F4320335322","funder_display_name":"European Regional Development Fund"},{"id":"https://openalex.org/G5421974903","display_name":null,"funder_award_id":"501100011033","funder_id":"https://openalex.org/F4320322930","funder_display_name":"Ministerio de Ciencia e Innovaci\u00f3n"},{"id":"https://openalex.org/G5852091828","display_name":null,"funder_award_id":"13039/501100011033","funder_id":"https://openalex.org/F4320334830","funder_display_name":"Ag\u00e8ncia de Gesti\u00f3 d'Ajuts Universitaris i de Recerca"},{"id":"https://openalex.org/G5967599077","display_name":null,"funder_award_id":"501100011033","funder_id":"https://openalex.org/F4320335322","funder_display_name":"European Regional Development Fund"},{"id":"https://openalex.org/G7084143925","display_name":null,"funder_award_id":"AEI/10","funder_id":"https://openalex.org/F4320335598","funder_display_name":"Agencia Estatal de Investigaci\u00f3n"}],"funders":[{"id":"https://openalex.org/F4320321505","display_name":"Generalitat de Catalunya","ror":"https://ror.org/01bg62x04"},{"id":"https://openalex.org/F4320322930","display_name":"Ministerio de Ciencia e Innovaci\u00f3n","ror":"https://ror.org/034900433"},{"id":"https://openalex.org/F4320334830","display_name":"Ag\u00e8ncia de Gesti\u00f3 d'Ajuts Universitaris i de Recerca","ror":"https://ror.org/01n4pqe45"},{"id":"https://openalex.org/F4320335322","display_name":"European Regional Development Fund","ror":"https://ror.org/00k4n6c32"},{"id":"https://openalex.org/F4320335598","display_name":"Agencia Estatal de Investigaci\u00f3n","ror":null}],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4413360549.pdf"},"referenced_works_count":6,"referenced_works":["https://openalex.org/W4321636575","https://openalex.org/W4385245566","https://openalex.org/W4387321091","https://openalex.org/W4389518760","https://openalex.org/W4394007410","https://openalex.org/W4409131852"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Large":[0],"language":[1,158],"models":[2],"have":[3],"been":[4],"widely":[5],"adopted":[6],"across":[7],"different":[8],"tasks,":[9],"but":[10],"their":[11],"auto-regressive":[12],"generation":[13],"nature":[14],"often":[15],"leads":[16],"to":[17,28,53,79,130],"inefficient":[18],"resource":[19,153],"utilization":[20],"during":[21],"inference.":[22],"While":[23],"batching":[24],"is":[25],"commonly":[26],"used":[27],"increase":[29],"throughput,":[30],"per-formance":[31],"gains":[32],"plateau":[33],"beyond":[34],"a":[35,43,51,92],"certain":[36],"batch":[37],"size,":[38],"especially":[39],"with":[40,72,105],"smaller":[41,157],"models,":[42],"phenomenon":[44],"that":[45,67,97],"existing":[46],"literature":[47],"typically":[48],"explains":[49],"as":[50,83],"shift":[52],"the":[54,84],"compute-bound":[55],"regime.":[56],"In":[57],"this":[58],"paper,":[59],"through":[60],"an":[61],"in-depth":[62],"GPU-level":[63],"analysis,":[64],"we":[65,90,126],"reveal":[66],"large-batch":[68],"inference":[69],"remains":[70],"memory-bound,":[71],"most":[73],"GPU":[74,102,115,135],"compute":[75,116],"capabilities":[76,117],"underutilized":[77,114],"due":[78],"DRAM":[80],"bandwidth":[81],"saturation":[82],"primary":[85],"bottleneck.":[86],"To":[87],"address":[88],"this,":[89],"propose":[91],"Batching":[93],"Configuration":[94],"Advisor":[95],"(BCA)":[96],"optimizes":[98],"memory":[99,103,112],"allocation,":[100],"reducing":[101],"requirements":[104],"minimal":[106],"impact":[107],"on":[108],"throughput.":[109],"The":[110],"freed":[111],"and":[113,134,148],"can":[118],"then":[119],"be":[120],"leveraged":[121],"by":[122],"concurrent":[123],"workloads.":[124],"Specifically,":[125],"use":[127],"model":[128],"replication":[129],"improve":[131],"serving":[132],"throughput":[133],"utilization.":[136],"Our":[137],"findings":[138],"challenge":[139],"conventional":[140],"assumptions":[141],"about":[142],"LLM":[143],"inference,":[144],"offering":[145],"new":[146],"in-sights":[147],"practical":[149],"strategies":[150],"for":[151,156],"improving":[152],"utilization,":[154],"particularly":[155],"models.":[159]},"counts_by_year":[{"year":2026,"cited_by_count":10},{"year":2025,"cited_by_count":3}],"updated_date":"2026-08-11T07:18:39.950985","created_date":"2025-10-10T00:00:00"}
