{"id":"https://openalex.org/W4411055324","doi":"https://doi.org/10.1109/mm.2025.3575361","title":"Splitwise: Efficient Generative LLM Inference Using Phase Splitting","display_name":"Splitwise: Efficient Generative LLM Inference Using Phase Splitting","publication_year":2025,"publication_date":"2025-06-05","ids":{"openalex":"https://openalex.org/W4411055324","doi":"https://doi.org/10.1109/mm.2025.3575361"},"language":"en","primary_location":{"id":"doi:10.1109/mm.2025.3575361","is_oa":false,"landing_page_url":"https://doi.org/10.1109/mm.2025.3575361","pdf_url":null,"source":{"id":"https://openalex.org/S59697426","display_name":"IEEE Micro","issn_l":"0272-1732","issn":["0272-1732","1937-4143"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Micro","raw_type":"journal-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/A5045708568","display_name":"Esha Choukse","orcid":"https://orcid.org/0000-0003-0371-5522"},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Esha Choukse","raw_affiliation_strings":["Microsoft Corporation, Redmond, WA, USA","Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA"],"raw_orcid":"https://orcid.org/0000-0003-0371-5522","affiliations":[{"raw_affiliation_string":"Microsoft Corporation, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]},{"raw_affiliation_string":"Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5073615366","display_name":"Pratyush Patel","orcid":"https://orcid.org/0000-0003-3611-5160"},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]},{"id":"https://openalex.org/I201448701","display_name":"University of Washington","ror":"https://ror.org/00cvxb145","country_code":"US","type":"education","lineage":["https://openalex.org/I201448701"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Pratyush Patel","raw_affiliation_strings":["University of Washington, Seattle, WA, USA","Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA"],"raw_orcid":"https://orcid.org/0000-0003-3611-5160","affiliations":[{"raw_affiliation_string":"University of Washington, Seattle, WA, USA","institution_ids":["https://openalex.org/I201448701"]},{"raw_affiliation_string":"Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045494456","display_name":"Chaojie Zhang","orcid":"https://orcid.org/0009-0002-8334-1291"},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chaojie Zhang","raw_affiliation_strings":["Microsoft Corporation, Redmond, WA, USA","Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA"],"raw_orcid":"https://orcid.org/0009-0002-8334-1291","affiliations":[{"raw_affiliation_string":"Microsoft Corporation, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]},{"raw_affiliation_string":"Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023546887","display_name":"Aashaka Shah","orcid":null},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Aashaka Shah","raw_affiliation_strings":["Microsoft Corporation, Redmond, WA, USA","Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA"],"raw_orcid":"https://orcid.org/0009-0004-0628-4515","affiliations":[{"raw_affiliation_string":"Microsoft Corporation, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]},{"raw_affiliation_string":"Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090311560","display_name":"\u00cd\u00f1igo Goiri","orcid":"https://orcid.org/0000-0003-2591-4012"},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"\u00cd\u00f1igo Goiri","raw_affiliation_strings":["Microsoft Corporation, Redmond, WA, USA","Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA"],"raw_orcid":"https://orcid.org/0000-0003-2591-4012","affiliations":[{"raw_affiliation_string":"Microsoft Corporation, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]},{"raw_affiliation_string":"Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Saeed Maleki","orcid":"https://orcid.org/0000-0002-7998-3681"},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]},{"id":"https://openalex.org/I173498003","display_name":"Palo Alto Research Center","ror":"https://ror.org/0529fxt39","country_code":"US","type":"facility","lineage":["https://openalex.org/I173498003","https://openalex.org/I4210132870"]},{"id":"https://openalex.org/I4210124598","display_name":"Palo Alto Institute","ror":"https://ror.org/02xf01n45","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I4210124598"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Saeed Maleki","raw_affiliation_strings":["xAI, Palo Alto, CA, USA","Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA"],"raw_orcid":"https://orcid.org/0000-0002-7998-3681","affiliations":[{"raw_affiliation_string":"xAI, Palo Alto, CA, USA","institution_ids":["https://openalex.org/I173498003","https://openalex.org/I4210124598"]},{"raw_affiliation_string":"Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036529548","display_name":"Rodrigo Fonseca","orcid":"https://orcid.org/0000-0001-9662-2661"},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Rodrigo Fonseca","raw_affiliation_strings":["Microsoft Corporation, Redmond, WA, USA","Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA"],"raw_orcid":"https://orcid.org/0000-0001-9662-2661","affiliations":[{"raw_affiliation_string":"Microsoft Corporation, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]},{"raw_affiliation_string":"Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5089615986","display_name":"Ricardo Bianchini","orcid":"https://orcid.org/0000-0001-5971-5084"},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ricardo Bianchini","raw_affiliation_strings":["Microsoft Corporation, Redmond, WA, USA","Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA"],"raw_orcid":"https://orcid.org/0000-0001-5971-5084","affiliations":[{"raw_affiliation_string":"Microsoft Corporation, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]},{"raw_affiliation_string":"Microsoft in the Azure Research \u2013 Systems group, Redmond, WA, USA","institution_ids":["https://openalex.org/I1290206253"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":8,"corresponding_author_ids":["https://openalex.org/A5045708568"],"corresponding_institution_ids":["https://openalex.org/I1290206253"],"apc_list":null,"apc_paid":null,"fwci":2.1998,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.87726298,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":96},"biblio":{"volume":"45","issue":"4","first_page":"54","last_page":"59"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T13717","display_name":"Advanced Algorithms and Applications","score":0.8165000081062317,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T13717","display_name":"Advanced Algorithms and Applications","score":0.8165000081062317,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.7407000064849854,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.7293999791145325,"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":0.8338807225227356},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6956278681755066},{"id":"https://openalex.org/keywords/generative-grammar","display_name":"Generative grammar","score":0.540911078453064},{"id":"https://openalex.org/keywords/phase","display_name":"Phase (matter)","score":0.4384712278842926},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.342602014541626},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.32613056898117065}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8338807225227356},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6956278681755066},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.540911078453064},{"id":"https://openalex.org/C44280652","wikidata":"https://www.wikidata.org/wiki/Q104837","display_name":"Phase (matter)","level":2,"score":0.4384712278842926},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.342602014541626},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.32613056898117065},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.0},{"id":"https://openalex.org/C178790620","wikidata":"https://www.wikidata.org/wiki/Q11351","display_name":"Organic chemistry","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/mm.2025.3575361","is_oa":false,"landing_page_url":"https://doi.org/10.1109/mm.2025.3575361","pdf_url":null,"source":{"id":"https://openalex.org/S59697426","display_name":"IEEE Micro","issn_l":"0272-1732","issn":["0272-1732","1937-4143"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Micro","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":1,"referenced_works":["https://openalex.org/W4387321091"],"related_works":["https://openalex.org/W2380075625","https://openalex.org/W4390718435","https://openalex.org/W2055243143","https://openalex.org/W4390549206","https://openalex.org/W3137171911","https://openalex.org/W4379540039","https://openalex.org/W4237784285","https://openalex.org/W2374712251","https://openalex.org/W4383031710","https://openalex.org/W3211753092"],"abstract_inverted_index":{"Generative":[0],"Large":[1],"Language":[2],"Model":[3],"(LLM)":[4],"applications":[5],"are":[6],"rapidly":[7],"growing,":[8],"leading":[9],"to":[10,97,117],"widespread":[11],"deployment":[12],"of":[13,20],"expensive,":[14],"power-hungry":[15],"GPUs.":[16],"Growing":[17],"power":[18,124],"demands":[19],"AI":[21],"in":[22],"the":[23,66,122],"cloud":[24],"industry":[25],"has":[26],"become":[27],"a":[28,42,47,85],"global":[29],"problem":[30],"[5].":[31],"Our":[32],"analysis":[33],"shows":[34],"that":[35,88],"LLM":[36],"inference":[37],"involves":[38],"two":[39],"distinct":[40],"phases:":[41],"compute-intensive":[43],"prefill":[44,64,90],"phase":[45,68],"and":[46,73,91,125,136],"memory-intensive":[48],"decode":[49,67,92],"phase,":[50,65],"each":[51],"with":[52],"different":[53,95],"resource":[54],"needs.":[55],"Running":[56],"them":[57],"together":[58],"introduces":[59],"inefficient":[60],"scheduling.":[61],"Furthermore,":[62],"unlike":[63],"can":[69],"run":[70],"on":[71,79],"lower-cost":[72],"lowerpower":[74],"hardware.":[75],"<p":[76],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[77],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">Building":[78],"these":[80],"insights,":[81],"we":[82],"propose":[83],"Splitwise,":[84],"scheduling":[86],"technique":[87],"splits":[89],"phases":[93],"across":[94],"machines":[96],"achieve":[98],"better":[99],"throughput.":[100],"Additionally,":[101],"Splitwise":[102,114],"allows":[103],"phase-specific":[104],"hardware":[105],"optimization.":[106],"By":[107],"efficiently":[108],"transferring":[109],"request":[110],"state":[111],"between":[112],"machines,":[113],"achieves":[115],"up":[116],"2.35\u00d7":[118],"more":[119],"throughput":[120,131],"within":[121],"same":[123,137],"cost":[126,135],"budgets,":[127],"or":[128],"1.4\u00d7":[129],"higher":[130],"at":[132],"20%":[133],"lower":[134],"power.":[138]},"counts_by_year":[{"year":2025,"cited_by_count":2}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}