{"id":"https://openalex.org/W7152673404","doi":"https://doi.org/10.48550/arxiv.2604.06970","title":"Scheduling the Unschedulable: Taming Black-Box LLM Inference at Scale","display_name":"Scheduling the Unschedulable: Taming Black-Box LLM Inference at Scale","publication_year":2026,"publication_date":"2026-04-08","ids":{"openalex":"https://openalex.org/W7152673404","doi":"https://doi.org/10.48550/arxiv.2604.06970"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.06970","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06970","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.2604.06970","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5103706195","display_name":"R.-Z. Yuan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuan, Renzhong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133312089","display_name":"Yijun Zeng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Yijun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133270737","display_name":"Xiaosong Gao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gao, Xiaosong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133302122","display_name":"Linxi Yu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Linxi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133255122","display_name":"Haochun Liao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liao, Haochun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133304434","display_name":"Han Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Han","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/T10101","display_name":"Cloud Computing and Resource Management","score":0.2184000015258789,"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"}},"topics":[{"id":"https://openalex.org/T10101","display_name":"Cloud Computing and Resource Management","score":0.2184000015258789,"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"}},{"id":"https://openalex.org/T12127","display_name":"Software System Performance and Reliability","score":0.2092999964952469,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T10714","display_name":"Software-Defined Networks and 5G","score":0.13349999487400055,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/goodput","display_name":"Goodput","score":0.7926999926567078},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.635200023651123},{"id":"https://openalex.org/keywords/queueing-theory","display_name":"Queueing theory","score":0.5819000005722046},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.5090000033378601},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.4449000060558319},{"id":"https://openalex.org/keywords/weighted-fair-queueing","display_name":"Weighted fair queueing","score":0.44339999556541443},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.38510000705718994},{"id":"https://openalex.org/keywords/thompson-sampling","display_name":"Thompson sampling","score":0.38040000200271606},{"id":"https://openalex.org/keywords/pooling","display_name":"Pooling","score":0.3792000114917755}],"concepts":[{"id":"https://openalex.org/C94022561","wikidata":"https://www.wikidata.org/wiki/Q1172393","display_name":"Goodput","level":4,"score":0.7926999926567078},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7828999757766724},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.635200023651123},{"id":"https://openalex.org/C22684755","wikidata":"https://www.wikidata.org/wiki/Q847526","display_name":"Queueing theory","level":2,"score":0.5819000005722046},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5090000033378601},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.4449000060558319},{"id":"https://openalex.org/C102486512","wikidata":"https://www.wikidata.org/wiki/Q457762","display_name":"Weighted fair queueing","level":3,"score":0.44339999556541443},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.4194999933242798},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.38510000705718994},{"id":"https://openalex.org/C73602740","wikidata":"https://www.wikidata.org/wiki/Q7795822","display_name":"Thompson sampling","level":3,"score":0.38040000200271606},{"id":"https://openalex.org/C70437156","wikidata":"https://www.wikidata.org/wiki/Q7228652","display_name":"Pooling","level":2,"score":0.3792000114917755},{"id":"https://openalex.org/C160403385","wikidata":"https://www.wikidata.org/wiki/Q220543","display_name":"Queue","level":2,"score":0.3644999861717224},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3610999882221222},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.35830000042915344},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.3285999894142151},{"id":"https://openalex.org/C195563490","wikidata":"https://www.wikidata.org/wiki/Q180368","display_name":"Network congestion","level":3,"score":0.32659998536109924},{"id":"https://openalex.org/C143647952","wikidata":"https://www.wikidata.org/wiki/Q1392979","display_name":"Fair queuing","level":5,"score":0.31700000166893005},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.31520000100135803},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.30809998512268066},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.29820001125335693},{"id":"https://openalex.org/C5119721","wikidata":"https://www.wikidata.org/wiki/Q220501","display_name":"Quality of service","level":2,"score":0.2980000078678131},{"id":"https://openalex.org/C75949130","wikidata":"https://www.wikidata.org/wiki/Q848010","display_name":"Database transaction","level":2,"score":0.2921999990940094},{"id":"https://openalex.org/C86127852","wikidata":"https://www.wikidata.org/wiki/Q7250364","display_name":"Proportionally fair","level":5,"score":0.27959999442100525},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.2793000042438507},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2782000005245209},{"id":"https://openalex.org/C2992947560","wikidata":"https://www.wikidata.org/wiki/Q5167661","display_name":"Fair share","level":2,"score":0.2678000032901764}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.06970","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06970","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.2604.06970","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06970","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":{"When":[0],"output":[1],"token":[2,27],"counts":[3],"can":[4],"be":[5],"predicted":[6],"at":[7],"submission":[8],"time":[9],"(Gan":[10],"et":[11],"al.,":[12],"2026),":[13],"client-side":[14,211],"scheduling":[15],"against":[16],"a":[17,63],"black-box":[18],"LLM":[19],"API":[20],"becomes":[21],"semi-clairvoyant:":[22],"decisions":[23],"condition":[24],"on":[25,62,151],"coarse":[26,72],"priors":[28,74],"even":[29],"though":[30],"the":[31,82,107,195,204,209,229],"provider's":[32],"internals":[33],"remain":[34],"hidden.":[35],"We":[36,157,207],"decompose":[37],"this":[38],"boundary":[39],"problem":[40],"into":[41],"three":[42],"separable":[43],"concerns:":[44],"allocation":[45,161,196],"(inter-class":[46],"share":[47],"via":[48],"adaptive":[49],"DRR),":[50],"ordering":[51],"(intra-class":[52],"sequencing":[53],"with":[54,125,166,181],"feasible-set":[55],"scoring),":[56],"and":[57,98,116,154,222],"overload":[58],"control":[59],"(explicit":[60],"admit/defer/reject":[61],"cost":[64],"ladder).":[65],"An":[66],"information":[67],"ladder":[68],"experiment":[69],"shows":[70],"that":[71,194],"magnitude":[73,90],"--":[75,80,192],"not":[76],"class":[77],"labels":[78],"alone":[79],"are":[81],"practical":[83],"threshold":[84],"for":[85],"useful":[86,117],"client":[87],"control;":[88],"removing":[89],"inflates":[91],"short-request":[92,176],"P95":[93,127],"by":[94],"up":[95,142],"to":[96,143,228],"$5.8\\times$":[97],"degrades":[99],"deadline":[100,114],"satisfaction.":[101],"Under":[102],"balanced":[103],"/":[104,189],"high":[105],"congestion":[106],"full":[108],"stack":[109],"achieves":[110,174],"100%":[111,113],"completion,":[112,152],"satisfaction,":[115],"goodput":[118],"of":[119,130,132,215],"$4.2":[120],"\\pm":[121],"1.6$":[122],"SLO-meeting":[123],"requests/s":[124],"short":[126],"within":[128],"tens":[129],"milliseconds":[131],"quota-tiered":[133],"isolation.":[134],"A":[135],"predictor-noise":[136],"sweep":[137],"confirms":[138],"graceful":[139],"degradation":[140],"under":[141],"60%":[144],"multiplicative":[145],"error.":[146],"Heavy-dominated":[147],"regimes":[148],"separate":[149],"policies":[150],"tail,":[153],"interpretable":[155],"shedding.":[156],"further":[158],"compare":[159],"short-priority":[160],"(biased":[162],"toward":[163],"interactive":[164],"traffic)":[165],"Fair":[167,172],"Queuing":[168,173],"(round-robin":[169],"across":[170,218],"classes):":[171],"+32%":[175],"P90":[177],"improvement":[178],"over":[179],"FIFO":[180],"only":[182],"+17%":[183],"long-request":[184],"overhead,":[185],"versus":[186],"Short-Priority's":[187],"+27%":[188],"+116%":[190],"trade-off":[191],"demonstrating":[193],"layer":[197],"accommodates":[198],"different":[199],"fairness":[200],"objectives":[201],"without":[202],"changing":[203],"remaining":[205],"stack.":[206],"contribute":[208],"three-layer":[210],"decomposition,":[212],"controlled":[213],"evaluation":[214],"joint":[216],"metrics":[217],"regimes,":[219],"allocation-policy":[220],"alternatives,":[221],"overload-policy":[223],"evidence":[224],"linking":[225],"cost-ladder":[226],"shedding":[227],"stated":[230],"service":[231],"objective.":[232]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-10T00:00:00"}