{"id":"https://openalex.org/W4414909300","doi":"https://doi.org/10.1109/cluster59342.2025.11186491","title":"SplitQuant: Resource-Efficient LLM Offline Serving on Heterogeneous GPUs via Phase-Aware Model Partition and Adaptive Quantization","display_name":"SplitQuant: Resource-Efficient LLM Offline Serving on Heterogeneous GPUs via Phase-Aware Model Partition and Adaptive Quantization","publication_year":2025,"publication_date":"2025-09-02","ids":{"openalex":"https://openalex.org/W4414909300","doi":"https://doi.org/10.1109/cluster59342.2025.11186491"},"language":"en","primary_location":{"id":"doi:10.1109/cluster59342.2025.11186491","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cluster59342.2025.11186491","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Cluster Computing (CLUSTER)","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":"https://openalex.org/A5110986442","display_name":"Juntao Zhao","orcid":null},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Juntao Zhao","raw_affiliation_strings":["The University of Hong Kong,Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Hong Kong,Hong Kong","institution_ids":["https://openalex.org/I177725633","https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5093967737","display_name":"Borui Wan","orcid":"https://orcid.org/0009-0008-5902-1611"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Borui Wan","raw_affiliation_strings":["The University of Hong Kong,Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Hong Kong,Hong Kong","institution_ids":["https://openalex.org/I177725633","https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013856812","display_name":"Yanghua Peng","orcid":"https://orcid.org/0000-0003-3989-4358"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yanghua Peng","raw_affiliation_strings":["ByteDance Inc.,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"ByteDance Inc.,USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084047386","display_name":"Haibin Lin","orcid":"https://orcid.org/0000-0003-4879-5335"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haibin Lin","raw_affiliation_strings":["ByteDance Inc.,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"ByteDance Inc.,USA","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101232348","display_name":"Chuan Wu","orcid":null},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Chuan Wu","raw_affiliation_strings":["The University of Hong Kong,Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Hong Kong,Hong Kong","institution_ids":["https://openalex.org/I177725633","https://openalex.org/I889458895"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.23234507,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"11"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9750000238418579,"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":0.9750000238418579,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.968500018119812,"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/T10052","display_name":"Medical Image Segmentation Techniques","score":0.9657999873161316,"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/quantization","display_name":"Quantization (signal processing)","score":0.73089998960495},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.6165000200271606},{"id":"https://openalex.org/keywords/partition","display_name":"Partition (number theory)","score":0.5888000130653381},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.4828999936580658},{"id":"https://openalex.org/keywords/sizing","display_name":"Sizing","score":0.4447999894618988},{"id":"https://openalex.org/keywords/homogeneous","display_name":"Homogeneous","score":0.41999998688697815},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.4020000100135803}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.798799991607666},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.73089998960495},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6165000200271606},{"id":"https://openalex.org/C42812","wikidata":"https://www.wikidata.org/wiki/Q1082910","display_name":"Partition (number theory)","level":2,"score":0.5888000130653381},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.5248000025749207},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.4828999936580658},{"id":"https://openalex.org/C2777767291","wikidata":"https://www.wikidata.org/wiki/Q1080291","display_name":"Sizing","level":2,"score":0.4447999894618988},{"id":"https://openalex.org/C66882249","wikidata":"https://www.wikidata.org/wiki/Q169336","display_name":"Homogeneous","level":2,"score":0.41999998688697815},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4146000146865845},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.4020000100135803},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.3977999985218048},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3278000056743622},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.2937999963760376},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2890999913215637},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.2851000130176544},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.2621999979019165},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.2581000030040741},{"id":"https://openalex.org/C81293917","wikidata":"https://www.wikidata.org/wiki/Q4189534","display_name":"System deployment","level":3,"score":0.25699999928474426},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.25519999861717224},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.25189998745918274},{"id":"https://openalex.org/C459310","wikidata":"https://www.wikidata.org/wiki/Q117801","display_name":"Computational science","level":1,"score":0.250900000333786}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/cluster59342.2025.11186491","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cluster59342.2025.11186491","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Cluster Computing (CLUSTER)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":15,"referenced_works":["https://openalex.org/W1632114991","https://openalex.org/W2058373514","https://openalex.org/W2606974598","https://openalex.org/W2963015836","https://openalex.org/W2982041622","https://openalex.org/W2991040477","https://openalex.org/W2998617917","https://openalex.org/W3094502228","https://openalex.org/W4231136072","https://openalex.org/W4321636575","https://openalex.org/W4387321091","https://openalex.org/W4402671258","https://openalex.org/W4404401018","https://openalex.org/W4407218342","https://openalex.org/W4408029577"],"related_works":[],"abstract_inverted_index":{"Modern":[0],"large":[1],"language":[2],"models":[3],"(LLMs)":[4],"serving":[5,88],"systems":[6,31,43],"address":[7],"distributed":[8,15,87],"deployment":[9],"challenges":[10],"through":[11],"two":[12,48],"key":[13],"techniques:":[14],"model":[16,95,122,159],"partitioning":[17,54,72,160],"for":[18,25,100,162],"parallel":[19],"computation":[20],"across":[21,126],"accelerators":[22,46],"and":[23,55,67,73,97,158],"quantization":[24,56,74,157],"reducing":[26],"parameter":[27],"size.":[28],"While":[29],"existing":[30],"assume":[32],"homogeneous":[33],"GPU":[34],"environments,":[35],"we":[36],"reveal":[37],"significant":[38],"untapped":[39],"potential":[40],"in":[41],"heterogeneous":[42,101,163],"with":[44,111],"mixed-capacity":[45],"where":[47],"critical":[49,76],"limitations":[50],"persist:":[51],"(1)":[52],"uniform":[53],"strategies":[57,161],"fail":[58],"to":[59,61,116,134],"adapt":[60],"hardware":[62],"heterogeneity,":[63],"exacerbating":[64],"resource":[65],"imbalance,":[66],"(2)":[68],"decoupled":[69],"optimization":[70],"of":[71,108,155],"overlooks":[75],"performance":[77],"synergies":[78],"between":[79],"these":[80],"techniques.":[81],"We":[82],"present":[83],"SplitQuant,":[84],"a":[85,112],"phase-aware":[86,94],"system":[89],"that":[90],"co-optimizes":[91],"mixedprecision":[92],"quantization,":[93],"partitioning,":[96],"micro-batch":[98],"sizing":[99],"environments.":[102,164],"Our":[103,150],"approach":[104],"combines":[105],"analytical":[106],"modeling":[107],"quality-runtime":[109],"tradeoffs":[110],"lightweight":[113],"planning":[114],"algorithm":[115],"maximize":[117],"throughput":[118,142],"while":[119],"preserving":[120],"user-specified":[121],"quality":[123],"targets.":[124,149],"Evaluations":[125],"10":[127],"production":[128],"clusters":[129],"show":[130],"SplitQuant":[131],"achieves":[132],"up":[133],"<tex":[135],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[136],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$2.34":[137],"\\times(1.61":[138],"\\times$</tex>":[139],"mean)":[140],"higher":[141],"than":[143],"state-of-theart":[144],"approaches":[145],"without":[146],"violating":[147],"accuracy":[148],"results":[151],"underscore":[152],"the":[153],"value":[154],"co-designing":[156]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}
