{"id":"https://openalex.org/W7163149329","doi":"https://doi.org/10.48550/arxiv.2606.01927","title":"Scaling LLM Inference Beyond Amdahl`s Limits via Eliminating Non-Scalable Overheads","display_name":"Scaling LLM Inference Beyond Amdahl`s Limits via Eliminating Non-Scalable Overheads","publication_year":2026,"publication_date":"2026-06-01","ids":{"openalex":"https://openalex.org/W7163149329","doi":"https://doi.org/10.48550/arxiv.2606.01927"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.01927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01927","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.2606.01927","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133047475","display_name":"Alan Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Alan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137698348","display_name":"Cyril Y. He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Cyril Y.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137669329","display_name":"Wei Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Wei","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":0.6254000067710876,"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":0.6254000067710876,"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/T10101","display_name":"Cloud Computing and Resource Management","score":0.1412999927997589,"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/T11181","display_name":"Advanced Data Storage Technologies","score":0.10379999876022339,"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":0.5909000039100647},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.5867999792098999},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5756000280380249},{"id":"https://openalex.org/keywords/degree","display_name":"Degree (music)","score":0.5001000165939331},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.4244999885559082},{"id":"https://openalex.org/keywords/efficient-energy-use","display_name":"Efficient energy use","score":0.41110000014305115},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.34689998626708984}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7319999933242798},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.5909000039100647},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.5867999792098999},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5756000280380249},{"id":"https://openalex.org/C2775997480","wikidata":"https://www.wikidata.org/wiki/Q586277","display_name":"Degree (music)","level":2,"score":0.5001000165939331},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.48489999771118164},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.47589999437332153},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.4244999885559082},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.41110000014305115},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.365200012922287},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.34689998626708984},{"id":"https://openalex.org/C2984822820","wikidata":"https://www.wikidata.org/wiki/Q1123036","display_name":"Processor scheduling","level":3,"score":0.3424000144004822},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.31700000166893005},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.30709999799728394},{"id":"https://openalex.org/C2989134064","wikidata":"https://www.wikidata.org/wiki/Q288510","display_name":"Execution time","level":2,"score":0.301800012588501},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.2973000109195709},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.28949999809265137},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.27410000562667847},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.26249998807907104},{"id":"https://openalex.org/C2780310539","wikidata":"https://www.wikidata.org/wiki/Q12547192","display_name":"Imperfect","level":2,"score":0.2547000050544739},{"id":"https://openalex.org/C157742956","wikidata":"https://www.wikidata.org/wiki/Q3237776","display_name":"Frequency scaling","level":3,"score":0.2522999942302704}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.01927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01927","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.2606.01927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01927","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":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Deployers":[0],"of":[1,15,93],"online":[2],"LLM":[3],"services":[4],"usually":[5],"seek":[6],"to":[7,22,36,113,135],"maximize":[8],"cluster-wide":[9],"performance":[10],"given":[11],"a":[12,77],"fixed":[13],"number":[14],"GPUs.":[16],"Tensor":[17],"parallelism":[18],"(TP)":[19],"is":[20],"necessary":[21],"fit":[23],"modern":[24],"models":[25,107],"but":[26],"scales":[27],"sub-linearly":[28],"as":[29,43],"the":[30,83,88],"TP":[31,67],"degree":[32,68],"t":[33,50],"grows,":[34],"due":[35],"cross-GPU":[37],"communication":[38],"and":[39,54,58,62,95,99,108,124],"non-scalable":[40,89],"runtime":[41],"work,":[42],"predicted":[44],"by":[45,86],"Amdahl's":[46],"Law.":[47],"Conversely,":[48],"increasing":[49],"improves":[51],"memory":[52],"efficiency":[53],"alleviates":[55],"KV-cache":[56],"contention":[57],"swapping.":[59],"We":[60,74],"identify":[61],"validate":[63],"an":[64],"empirical":[65],"optimal":[66],"t_e":[69,85],"that":[70,81],"balances":[71],"these":[72],"effects.":[73],"present":[75],"Albireo,":[76],"parallel":[78],"inference":[79],"system":[80],"raises":[82],"attainable":[84],"shrinking":[87],"portion":[90],"via":[91],"overlap":[92],"scheduling":[94],"I/O":[96],"with":[97],"compute":[98],"sequence-parallel":[100],"sampling,":[101],"without":[102],"changing":[103],"model":[104],"architectures.":[105],"Across":[106],"benchmarks,":[109],"Albireo":[110],"achieves":[111],"up":[112,134],"1.9x":[114],"higher":[115,121,137],"throughput,":[116],"48%":[117],"lower":[118,126],"latency,":[119],"28%":[120],"GPU":[122],"utilization,":[123],"54%":[125],"energy":[127],"than":[128],"vLLM;":[129],"in":[130],"production":[131],"it":[132],"yields":[133],"2x":[136],"throughput.":[138]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-03T00:00:00"}