{"id":"https://openalex.org/W7124173814","doi":"https://doi.org/10.1109/icpads67057.2025.11323175","title":"MoEoM: Joint Compute and Memory-Aware Balancing for Fast MoE Inference","display_name":"MoEoM: Joint Compute and Memory-Aware Balancing for Fast MoE Inference","publication_year":2025,"publication_date":"2025-12-14","ids":{"openalex":"https://openalex.org/W7124173814","doi":"https://doi.org/10.1109/icpads67057.2025.11323175"},"language":null,"primary_location":{"id":"doi:10.1109/icpads67057.2025.11323175","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icpads67057.2025.11323175","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 31th International Conference on Parallel and Distributed Systems (ICPADS)","raw_type":"proceedings-article"},"type":"conference-paper","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/A5123038748","display_name":"Ziqi Gong","orcid":null},"institutions":[{"id":"https://openalex.org/I162868743","display_name":"Tianjin University","ror":"https://ror.org/012tb2g32","country_code":"CN","type":"education","lineage":["https://openalex.org/I162868743"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ziqi Gong","raw_affiliation_strings":["Tianjin University,Tianjin Key Laboratory of Advanced Networking,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tianjin University,Tianjin Key Laboratory of Advanced Networking,China","institution_ids":["https://openalex.org/I162868743"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5122802787","display_name":"Yitao Hu","orcid":null},"institutions":[{"id":"https://openalex.org/I162868743","display_name":"Tianjin University","ror":"https://ror.org/012tb2g32","country_code":"CN","type":"education","lineage":["https://openalex.org/I162868743"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yitao Hu","raw_affiliation_strings":["Tianjin University,Tianjin Key Laboratory of Advanced Networking,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tianjin University,Tianjin Key Laboratory of Advanced Networking,China","institution_ids":["https://openalex.org/I162868743"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100320971","display_name":"Sheng Chen","orcid":"https://orcid.org/0000-0001-7038-4407"},"institutions":[{"id":"https://openalex.org/I162868743","display_name":"Tianjin University","ror":"https://ror.org/012tb2g32","country_code":"CN","type":"education","lineage":["https://openalex.org/I162868743"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Sheng Chen","raw_affiliation_strings":["Tianjin University,Tianjin Key Laboratory of Advanced Networking,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tianjin University,Tianjin Key Laboratory of Advanced Networking,China","institution_ids":["https://openalex.org/I162868743"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5122986283","display_name":"Wenxin Li","orcid":null},"institutions":[{"id":"https://openalex.org/I162868743","display_name":"Tianjin University","ror":"https://ror.org/012tb2g32","country_code":"CN","type":"education","lineage":["https://openalex.org/I162868743"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wenxin Li","raw_affiliation_strings":["Tianjin University,Tianjin Key Laboratory of Advanced Networking,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tianjin University,Tianjin Key Laboratory of Advanced Networking,China","institution_ids":["https://openalex.org/I162868743"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5111982109","display_name":"Keqiu Li","orcid":"https://orcid.org/0000-0003-1758-3030"},"institutions":[{"id":"https://openalex.org/I162868743","display_name":"Tianjin University","ror":"https://ror.org/012tb2g32","country_code":"CN","type":"education","lineage":["https://openalex.org/I162868743"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Keqiu Li","raw_affiliation_strings":["Tianjin University,Tianjin Key Laboratory of Advanced Networking,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tianjin University,Tianjin Key Laboratory of Advanced Networking,China","institution_ids":["https://openalex.org/I162868743"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I162868743"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.57217549,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"10"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.5199000239372253,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.5199000239372253,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.06530000269412994,"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/T14347","display_name":"Big Data and Digital Economy","score":0.03819999843835831,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.7483999729156494},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.6245999932289124},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5796999931335449},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.5695000290870667},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.5407000184059143},{"id":"https://openalex.org/keywords/backup","display_name":"Backup","score":0.5196999907493591},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.3896999955177307},{"id":"https://openalex.org/keywords/offset","display_name":"Offset (computer science)","score":0.3736000061035156}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7620000243186951},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7483999729156494},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.6245999932289124},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5796999931335449},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.5695000290870667},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.5407000184059143},{"id":"https://openalex.org/C2780945871","wikidata":"https://www.wikidata.org/wiki/Q194274","display_name":"Backup","level":2,"score":0.5196999907493591},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.4189999997615814},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3896999955177307},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.3736000061035156},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3353999853134155},{"id":"https://openalex.org/C55282118","wikidata":"https://www.wikidata.org/wiki/Q252683","display_name":"Snapshot (computer storage)","level":2,"score":0.32670000195503235},{"id":"https://openalex.org/C18555067","wikidata":"https://www.wikidata.org/wiki/Q8375051","display_name":"Joint (building)","level":2,"score":0.30489999055862427},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.30300000309944153},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.28690001368522644},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.2827000021934509},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2797999978065491},{"id":"https://openalex.org/C138959212","wikidata":"https://www.wikidata.org/wiki/Q1806783","display_name":"Load balancing (electrical power)","level":3,"score":0.27790001034736633},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.27559998631477356},{"id":"https://openalex.org/C2777655017","wikidata":"https://www.wikidata.org/wiki/Q1501161","display_name":"Toolbox","level":2,"score":0.2653999924659729},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.2637999951839447},{"id":"https://openalex.org/C93996380","wikidata":"https://www.wikidata.org/wiki/Q44127","display_name":"Server","level":2,"score":0.2583000063896179}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icpads67057.2025.11323175","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icpads67057.2025.11323175","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 31th International Conference on Parallel and Distributed Systems (ICPADS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G5907215754","display_name":null,"funder_award_id":"62572341,62202328","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G6975830235","display_name":null,"funder_award_id":"2022YFB4501000","funder_id":"https://openalex.org/F4320335777","funder_display_name":"National Key Research and Development Program of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320335777","display_name":"National Key Research and Development Program of China","ror":null}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":4,"referenced_works":["https://openalex.org/W1988645907","https://openalex.org/W2150884987","https://openalex.org/W4385245566","https://openalex.org/W4387321091"],"related_works":[],"abstract_inverted_index":{"Mixture-of-Experts":[0],"(MoE)":[1],"architectures":[2],"have":[3],"emerged":[4],"as":[5],"a":[6,18,100],"scalable":[7],"and":[8,39,63,111,116,129,166,173,200,206],"efficient":[9,210],"alternative":[10],"to":[11,35,50,52,132,154,186],"dense":[12],"Transformer":[13],"models":[14,27,86,144],"by":[15,87,152,159,177],"activating":[16],"only":[17],"subset":[19],"of":[20,70,84,196],"experts":[21,108,131],"per":[22],"layer.":[23],"However,":[24],"deploying":[25],"MoE":[26,85,143,188,212],"in":[28,162,169],"multi-GPU":[29],"environments":[30],"faces":[31],"severe":[32],"challenges":[33],"due":[34],"expert":[36,103,204],"load":[37,71],"imbalance":[38,176],"the":[40,68,82,134,163,170,194],"resulting":[41],"inefficient":[42],"GPU":[43],"utilization.":[44],"Existing":[45],"static":[46],"replication":[47],"strategies":[48],"fail":[49],"adapt":[51],"dynamic":[53,57],"token":[54,121,207],"distributions,":[55],"while":[56],"rebalancing":[58],"methods":[59],"incur":[60],"excessive":[61],"communication":[62],"memory-access":[64,90,201],"overheads,":[65],"often":[66],"outweighing":[67],"benefits":[69],"balancing.":[72],"This":[73],"paper":[74],"presents":[75],"MoEoM,":[76],"an":[77,118],"inference":[78,150,189],"system":[79],"that":[80,146],"enhances":[81],"efficiency":[83],"innovatively":[88],"taking":[89],"costs":[91],"into":[92,203],"consideration.":[93],"MoEoM":[94,147],"integrates":[95],"two":[96],"complementary":[97],"techniques:":[98],"(i)":[99],"load-aware":[101],"offline":[102],"deployer,":[104],"which":[105,123],"symmetrically":[106],"groups":[107],"across":[109,137],"GPUs":[110],"selectively":[112],"replicates":[113],"high-load":[114],"experts,":[115],"(ii)":[117],"I/O-aware":[119],"online":[120],"reallocator,":[122],"dynamically":[124],"redistributes":[125],"tokens":[126],"among":[127],"original":[128],"backup":[130],"minimize":[133],"maximum":[135],"latency":[136,151],"GPUs.":[138],"Experimental":[139],"evaluations":[140],"on":[141,183],"state-of-theart":[142],"demonstrate":[145],"reduces":[148],"end-toend":[149],"up":[153],"16.6":[155],"%,":[156],"improves":[157],"throughput":[158],"11\u201320":[160],"%":[161,168,179,182],"prefill":[164],"stage":[165],"11\u201323":[167],"decode":[171],"stage,":[172],"decreases":[174],"cross-GPU":[175],"65\u201380":[178],"(about":[180],"75":[181],"average),":[184],"compared":[185],"prior":[187],"baselines.":[190],"These":[191],"results":[192],"highlight":[193],"importance":[195],"incorporating":[197],"both":[198],"computation":[199],"overhead":[202],"placement":[205],"scheduling":[208],"for":[209],"large-scale":[211],"deployment.":[213]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-01-15T00:00:00"}