{"id":"https://openalex.org/W7164495321","doi":"https://doi.org/10.48550/arxiv.2606.13501","title":"GF-DiT: Scheduling Parallelism for Diffusion Transformer Serving","display_name":"GF-DiT: Scheduling Parallelism for Diffusion Transformer Serving","publication_year":2026,"publication_date":"2026-06-11","ids":{"openalex":"https://openalex.org/W7164495321","doi":"https://doi.org/10.48550/arxiv.2606.13501"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.13501","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.13501","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.13501","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5119806774","display_name":"Xinwei Qiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qiang, Xinwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138485743","display_name":"Yifan Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Yifan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043164368","display_name":"Shixuan Sun","orcid":"https://orcid.org/0000-0003-4060-9438"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Shixuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100404947","display_name":"Jie Yang","orcid":"https://orcid.org/0000-0003-4801-7162"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Jing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003608372","display_name":"Hongyun Zhao","orcid":"https://orcid.org/0000-0002-1251-3888"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Han","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138530286","display_name":"Chen Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138543265","display_name":"Yu Feng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Feng, Yu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003939279","display_name":"Jingwen Leng","orcid":"https://orcid.org/0000-0002-5660-5493"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leng, Jingwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5138510316","display_name":"Minyi Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Minyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.2467000037431717,"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"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.2467000037431717,"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/T11019","display_name":"Image Enhancement Techniques","score":0.23399999737739563,"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/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.08070000261068344,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.555899977684021},{"id":"https://openalex.org/keywords/asynchronous-communication","display_name":"Asynchronous communication","score":0.5458999872207642},{"id":"https://openalex.org/keywords/workload","display_name":"Workload","score":0.5347999930381775},{"id":"https://openalex.org/keywords/control-reconfiguration","display_name":"Control reconfiguration","score":0.5066999793052673},{"id":"https://openalex.org/keywords/abstraction","display_name":"Abstraction","score":0.3804999887943268},{"id":"https://openalex.org/keywords/skew","display_name":"Skew","score":0.3797000050544739},{"id":"https://openalex.org/keywords/task-parallelism","display_name":"Task parallelism","score":0.34360000491142273},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.33570000529289246}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.864300012588501},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.555899977684021},{"id":"https://openalex.org/C151319957","wikidata":"https://www.wikidata.org/wiki/Q752739","display_name":"Asynchronous communication","level":2,"score":0.5458999872207642},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5424000024795532},{"id":"https://openalex.org/C2778476105","wikidata":"https://www.wikidata.org/wiki/Q628539","display_name":"Workload","level":2,"score":0.5347999930381775},{"id":"https://openalex.org/C119701452","wikidata":"https://www.wikidata.org/wiki/Q5165881","display_name":"Control reconfiguration","level":2,"score":0.5066999793052673},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.41690000891685486},{"id":"https://openalex.org/C124304363","wikidata":"https://www.wikidata.org/wiki/Q673661","display_name":"Abstraction","level":2,"score":0.3804999887943268},{"id":"https://openalex.org/C43711488","wikidata":"https://www.wikidata.org/wiki/Q7534783","display_name":"Skew","level":2,"score":0.3797000050544739},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.3553999960422516},{"id":"https://openalex.org/C42992933","wikidata":"https://www.wikidata.org/wiki/Q691169","display_name":"Task parallelism","level":3,"score":0.34360000491142273},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.33570000529289246},{"id":"https://openalex.org/C61483411","wikidata":"https://www.wikidata.org/wiki/Q3124522","display_name":"Data parallelism","level":3,"score":0.3285999894142151},{"id":"https://openalex.org/C53833338","wikidata":"https://www.wikidata.org/wiki/Q1061424","display_name":"Context switch","level":2,"score":0.3199999928474426},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.31779998540878296},{"id":"https://openalex.org/C193702766","wikidata":"https://www.wikidata.org/wiki/Q1414548","display_name":"Concurrency","level":2,"score":0.3131999969482422},{"id":"https://openalex.org/C2989134064","wikidata":"https://www.wikidata.org/wiki/Q288510","display_name":"Execution time","level":2,"score":0.3050000071525574},{"id":"https://openalex.org/C31395832","wikidata":"https://www.wikidata.org/wiki/Q1318674","display_name":"Testbed","level":2,"score":0.30489999055862427},{"id":"https://openalex.org/C2776834041","wikidata":"https://www.wikidata.org/wiki/Q25346349","display_name":"Execution model","level":2,"score":0.2921000123023987},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.2782999873161316},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.27309998869895935},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2685999870300293},{"id":"https://openalex.org/C169590947","wikidata":"https://www.wikidata.org/wiki/Q47506","display_name":"Compiler","level":2,"score":0.2671999931335449}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.13501","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.13501","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.13501","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.13501","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Diffusion":[0],"Transformers":[1],"(DiTs)":[2],"have":[3],"become":[4],"the":[5,88],"dominant":[6],"architecture":[7],"for":[8,16,81],"image":[9,154],"and":[10,42,49,56,97,114,138,149,155,188],"video":[11,156],"generation,":[12],"creating":[13],"growing":[14],"demand":[15],"efficient":[17],"DiT":[18,33,64,83],"serving.":[19],"Existing":[20],"systems":[21],"assign":[22],"each":[23],"request":[24],"a":[25,71,78,129],"fixed":[26],"parallel":[27],"configuration":[28],"throughout":[29],"its":[30],"lifetime.":[31],"However,":[32],"workloads":[34],"exhibit":[35],"substantial":[36],"heterogeneity":[37],"across":[38],"requests,":[39],"execution":[40,104,142,162],"stages,":[41],"system":[43],"conditions,":[44],"making":[45],"static":[46,164],"parallelism":[47,69,89,122],"inefficient":[48],"often":[50],"leading":[51],"to":[52,94,171,178,186,196],"poor":[53],"GPU":[54,68,117],"utilization":[55],"degraded":[57],"service":[58,98],"quality.":[59],"This":[60],"paper":[61],"argues":[62],"that":[63,85,106,133],"serving":[65,84],"should":[66],"treat":[67],"as":[70],"first-class":[72],"schedulable":[73,111],"resource.":[74],"We":[75,144],"present":[76],"GF-DiT,":[77],"policy-programmable":[79],"runtime":[80],"elastic":[82,121],"dynamically":[86],"adapts":[87],"of":[90,140],"running":[91],"requests":[92,108],"according":[93],"workload":[95],"demands":[96],"objectives.":[99],"GF-DiT":[100,124,146,166],"introduces":[101],"an":[102],"asynchronous":[103],"abstraction":[105,132],"decomposes":[107],"into":[109],"independently":[110],"trajectory":[112],"tasks":[113],"enables":[115],"online":[116,136],"reallocation.":[118],"To":[119],"make":[120],"practical,":[123],"further":[125],"proposes":[126],"group-free":[127],"collectives,":[128],"lightweight":[130],"communication":[131],"supports":[134],"low-overhead":[135],"formation":[137],"reconfiguration":[139],"arbitrary":[141],"groups.":[143],"implement":[145],"in":[147],"vLLM-Omni":[148],"evaluate":[150],"it":[151],"on":[152],"representative":[153],"diffusion":[157],"workloads.":[158],"Compared":[159],"with":[160,163],"fixed-pipeline":[161],"parallelism,":[165],"improves":[167],"throughput":[168],"by":[169,176,184],"up":[170,177,185],"6.01$\\times$,":[172],"reduces":[173,189],"mean":[174],"latency":[175],"95%,":[179],"lowers":[180],"SLO":[181],"violation":[182],"rates":[183],"90%,":[187],"communication-group":[190],"setup":[191],"overhead":[192],"from":[193],"778":[194],"ms":[195],"approximately":[197],"60":[198],"$\u03bc$s.":[199],"Our":[200],"code":[201],"is":[202],"available":[203],"at":[204],"https://github.com/SJTU-Liquid/GF-DiT.":[205]},"counts_by_year":[],"updated_date":"2026-07-04T06:09:54.619538","created_date":"2026-06-13T00:00:00"}
