{"id":"https://openalex.org/W7164515296","doi":"https://doi.org/10.48550/arxiv.2606.12950","title":"Maestro: Workload-Aware Cross-Cluster Scheduling for LLM-Based Multi-Agent Systems","display_name":"Maestro: Workload-Aware Cross-Cluster Scheduling for LLM-Based Multi-Agent Systems","publication_year":2026,"publication_date":"2026-06-11","ids":{"openalex":"https://openalex.org/W7164515296","doi":"https://doi.org/10.48550/arxiv.2606.12950"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.12950","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.12950","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":null,"license_id":null,"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.2606.12950","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5115028374","display_name":"Jinghao Wang","orcid":"https://orcid.org/0009-0009-5919-5328"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Jinghao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113903386","display_name":"X J Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Xiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138532422","display_name":"Xiaoyang Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Xiaoyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138552974","display_name":"Yihui Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yihui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138541085","display_name":"Yilong Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Yilong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066123431","display_name":"Tianyu Wo","orcid":"https://orcid.org/0000-0002-5331-3364"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wo, Tianyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138547253","display_name":"Xu Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138513174","display_name":"Chunming Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Chunming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5138550460","display_name":"Renyu Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Renyu","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/T14347","display_name":"Big Data and Digital Economy","score":0.3237999975681305,"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/T14347","display_name":"Big Data and Digital Economy","score":0.3237999975681305,"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/T10101","display_name":"Cloud Computing and Resource Management","score":0.09510000050067902,"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/T10273","display_name":"IoT and Edge/Fog Computing","score":0.07699999958276749,"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/scheduling","display_name":"Scheduling (production processes)","score":0.6256999969482422},{"id":"https://openalex.org/keywords/cloud-computing","display_name":"Cloud computing","score":0.5102999806404114},{"id":"https://openalex.org/keywords/workflow","display_name":"Workflow","score":0.49000000953674316},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.36660000681877136},{"id":"https://openalex.org/keywords/dynamic-priority-scheduling","display_name":"Dynamic priority scheduling","score":0.3596000075340271},{"id":"https://openalex.org/keywords/shared-resource","display_name":"Shared resource","score":0.3206999897956848},{"id":"https://openalex.org/keywords/response-time","display_name":"Response time","score":0.3190000057220459}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8759999871253967},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.6657999753952026},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.6256999969482422},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.5102999806404114},{"id":"https://openalex.org/C177212765","wikidata":"https://www.wikidata.org/wiki/Q627335","display_name":"Workflow","level":2,"score":0.49000000953674316},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.36660000681877136},{"id":"https://openalex.org/C107568181","wikidata":"https://www.wikidata.org/wiki/Q5319000","display_name":"Dynamic priority scheduling","level":3,"score":0.3596000075340271},{"id":"https://openalex.org/C51332947","wikidata":"https://www.wikidata.org/wiki/Q1172305","display_name":"Shared resource","level":2,"score":0.3206999897956848},{"id":"https://openalex.org/C19012869","wikidata":"https://www.wikidata.org/wiki/Q578372","display_name":"Response time","level":2,"score":0.3190000057220459},{"id":"https://openalex.org/C55416958","wikidata":"https://www.wikidata.org/wiki/Q6206757","display_name":"Job shop scheduling","level":3,"score":0.29120001196861267},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.29019999504089355},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.2759999930858612},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.27230000495910645},{"id":"https://openalex.org/C113364801","wikidata":"https://www.wikidata.org/wiki/Q26674","display_name":"High fidelity","level":2,"score":0.26589998602867126},{"id":"https://openalex.org/C169590947","wikidata":"https://www.wikidata.org/wiki/Q47506","display_name":"Compiler","level":2,"score":0.25760000944137573},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.25189998745918274},{"id":"https://openalex.org/C175893541","wikidata":"https://www.wikidata.org/wiki/Q1196582","display_name":"Round-robin scheduling","level":4,"score":0.25060001015663147}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.12950","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.12950","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.12950","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.12950","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"display_name":"Decent work and economic growth","id":"https://metadata.un.org/sdg/8","score":0.6384885311126709}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Model":[2],"based":[3],"Multi-Agent":[4],"Systems":[5],"(LLM-MAS)":[6],"have":[7],"emerged":[8],"as":[9],"a":[10,85,121],"powerful":[11],"paradigm":[12],"for":[13,90,168],"tackling":[14],"complex":[15],"tasks":[16],"by":[17,181,190],"breaking":[18],"them":[19],"into":[20],"collaborative":[21],"workflows":[22,61],"of":[23,45,112],"specialized":[24],"LLM-powered":[25],"agents.":[26],"However,":[27],"deploying":[28],"such":[29],"multi-agent":[30],"workloads":[31],"at":[32,67],"scale":[33],"poses":[34],"significant":[35],"system":[36,88],"challenges.":[37],"Each":[38],"user":[39],"query":[40],"spawns":[41],"an":[42],"iterative":[43],"pipeline":[44],"LLM":[46],"calls,":[47],"greatly":[48],"amplifying":[49],"resource":[50],"consumption":[51],"compared":[52],"to":[53,119,149,164],"single-turn":[54],"queries.":[55],"In":[56],"resource-constrained":[57],"cloud":[58],"settings,":[59],"these":[60],"face":[62],"non-deterministic":[63],"and":[64,76,78,102,109,115,137,153,174,183],"input-dependent":[65],"costs":[66],"decode":[68],"stage,":[69],"heavy-tailed":[70],"multi-model":[71,131],"requirements":[72],"with":[73],"memory":[74,110,139,154],"fragmentation":[75],"over-provisioning,":[77],"cross-cluster":[79],"scheduling":[80,87],"trade-offs.":[81],"We":[82],"present":[83],"Maestro,":[84],"workload-aware":[86],"designed":[89],"LLM-MAS":[91],"serving":[92],"under":[93],"strict":[94],"GPU":[95],"budgets.":[96],"Maestro":[97,128,177],"explicitly":[98],"leverages":[99],"agent":[100],"semantics":[101],"roles:":[103],"it":[104,145,160],"predicts":[105],"the":[106,125,142,157],"output":[107],"length":[108],"usage":[111],"each":[113],"stage":[114],"uses":[116],"this":[117],"prediction":[118],"drive":[120],"hierarchical":[122,134],"scheduler.":[123],"At":[124,141,156],"node":[126],"level,":[127,144,159],"enables":[129],"dynamic":[130],"co-location":[132],"via":[133],"weight":[135],"caching":[136],"elastic":[138],"provisioning.":[140],"cluster":[143],"performs":[146],"latency-aware":[147],"routing":[148],"avoid":[150],"cold-start":[151],"delays":[152],"overloads.":[155],"global":[158],"enforces":[161],"workflow-aware":[162],"prioritization":[163],"minimize":[165],"head-of-line":[166],"blocking":[167],"interactive":[169],"tasks.":[170],"Across":[171],"prototype":[172],"experiments":[173],"trace-driven":[175],"simulations,":[176],"reduces":[178],"KV-reservation":[179],"HBM":[180],"67.2%":[182],"improves":[184],"high-contention":[185],"SLO":[186],"attainment":[187],"over":[188],"EDF":[189],"23.6":[191],"percentage":[192],"points.":[193]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-13T00:00:00"}