{"id":"https://openalex.org/W2740001873","doi":"https://doi.org/10.1145/3236367.3236381","title":"Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters","display_name":"Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters","publication_year":2018,"publication_date":"2018-09-19","ids":{"openalex":"https://openalex.org/W2740001873","doi":"https://doi.org/10.1145/3236367.3236381","mag":"2740001873"},"language":"en","primary_location":{"id":"doi:10.1145/3236367.3236381","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3236367.3236381","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 25th European MPI Users' Group Meeting","raw_type":"proceedings-article"},"type":"article","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/A5004330728","display_name":"Ammar Ahmad Awan","orcid":"https://orcid.org/0000-0002-6272-3760"},"institutions":[{"id":"https://openalex.org/I52357470","display_name":"The Ohio State University","ror":"https://ror.org/00rs6vg23","country_code":"US","type":"education","lineage":["https://openalex.org/I52357470"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Ammar Ahmad Awan","raw_affiliation_strings":["Department of Computer Science and Engineering, The Ohio State University, Columbus, Ohio"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering, The Ohio State University, Columbus, Ohio","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076405830","display_name":"Ching-Hsiang Chu","orcid":"https://orcid.org/0000-0002-6752-3135"},"institutions":[{"id":"https://openalex.org/I52357470","display_name":"The Ohio State University","ror":"https://ror.org/00rs6vg23","country_code":"US","type":"education","lineage":["https://openalex.org/I52357470"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ching-Hsiang Chu","raw_affiliation_strings":["Department of Computer Science and Engineering, The Ohio State University, Columbus, Ohio"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering, The Ohio State University, Columbus, Ohio","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034293705","display_name":"Hari Subramoni","orcid":"https://orcid.org/0000-0002-1200-2754"},"institutions":[{"id":"https://openalex.org/I52357470","display_name":"The Ohio State University","ror":"https://ror.org/00rs6vg23","country_code":"US","type":"education","lineage":["https://openalex.org/I52357470"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hari Subramoni","raw_affiliation_strings":["Department of Computer Science and Engineering, The Ohio State University, Columbus, Ohio"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering, The Ohio State University, Columbus, Ohio","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5024879682","display_name":"Dhabaleswar K. Panda","orcid":"https://orcid.org/0000-0002-0356-1781"},"institutions":[{"id":"https://openalex.org/I52357470","display_name":"The Ohio State University","ror":"https://ror.org/00rs6vg23","country_code":"US","type":"education","lineage":["https://openalex.org/I52357470"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Dhabaleswar K. Panda","raw_affiliation_strings":["Department of Computer Science and Engineering, The Ohio State University, Columbus, Ohio"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering, The Ohio State University, Columbus, Ohio","institution_ids":["https://openalex.org/I52357470"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5004330728"],"corresponding_institution_ids":["https://openalex.org/I52357470"],"apc_list":null,"apc_paid":null,"fwci":2.6021,"has_fulltext":false,"cited_by_count":36,"citation_normalized_percentile":{"value":0.90115712,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"9"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9980000257492065,"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"}},"topics":[{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9980000257492065,"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"}},{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9966999888420105,"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/T11478","display_name":"Caching and Content Delivery","score":0.9952999949455261,"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/computer-science","display_name":"Computer science","score":0.8898985385894775},{"id":"https://openalex.org/keywords/infiniband","display_name":"InfiniBand","score":0.8616026043891907},{"id":"https://openalex.org/keywords/cuda","display_name":"CUDA","score":0.6377716064453125},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5841103792190552},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.5649964213371277},{"id":"https://openalex.org/keywords/low-latency","display_name":"Low latency (capital markets)","score":0.5200809836387634},{"id":"https://openalex.org/keywords/gpu-cluster","display_name":"GPU cluster","score":0.48617738485336304},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.4735647439956665},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.44378662109375},{"id":"https://openalex.org/keywords/distributed-computing","display_name":"Distributed computing","score":0.33588844537734985},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.15243202447891235},{"id":"https://openalex.org/keywords/telecommunications","display_name":"Telecommunications","score":0.08540675044059753}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8898985385894775},{"id":"https://openalex.org/C2781030343","wikidata":"https://www.wikidata.org/wiki/Q922437","display_name":"InfiniBand","level":2,"score":0.8616026043891907},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.6377716064453125},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5841103792190552},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.5649964213371277},{"id":"https://openalex.org/C46637626","wikidata":"https://www.wikidata.org/wiki/Q6693015","display_name":"Low latency (capital markets)","level":2,"score":0.5200809836387634},{"id":"https://openalex.org/C2781335571","wikidata":"https://www.wikidata.org/wiki/Q2633544","display_name":"GPU cluster","level":3,"score":0.48617738485336304},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.4735647439956665},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.44378662109375},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.33588844537734985},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.15243202447891235},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.08540675044059753},{"id":"https://openalex.org/C151730666","wikidata":"https://www.wikidata.org/wiki/Q7205","display_name":"Paleontology","level":1,"score":0.0},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3236367.3236381","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3236367.3236381","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 25th European MPI Users' Group Meeting","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Industry, innovation and infrastructure","id":"https://metadata.un.org/sdg/9","score":0.5099999904632568}],"awards":[],"funders":[{"id":"https://openalex.org/F4320309480","display_name":"Nvidia","ror":"https://ror.org/03jdj4y14"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":31,"referenced_works":["https://openalex.org/W1576599451","https://openalex.org/W1577304665","https://openalex.org/W1585237817","https://openalex.org/W1686810756","https://openalex.org/W1964981582","https://openalex.org/W2042670967","https://openalex.org/W2076063813","https://openalex.org/W2105549957","https://openalex.org/W2126852285","https://openalex.org/W2131613942","https://openalex.org/W2131940306","https://openalex.org/W2146375161","https://openalex.org/W2151194859","https://openalex.org/W2155893237","https://openalex.org/W2167967054","https://openalex.org/W2168231600","https://openalex.org/W2170135819","https://openalex.org/W2294872507","https://openalex.org/W2535374105","https://openalex.org/W2580688187","https://openalex.org/W2581372641","https://openalex.org/W2592864539","https://openalex.org/W2618530766","https://openalex.org/W2919115771","https://openalex.org/W2962801832","https://openalex.org/W2962911728","https://openalex.org/W3105781365","https://openalex.org/W4233199365","https://openalex.org/W4256078451","https://openalex.org/W4285719527","https://openalex.org/W6829887170"],"related_works":["https://openalex.org/W1449833061","https://openalex.org/W2030707850","https://openalex.org/W2161462353","https://openalex.org/W3101208208","https://openalex.org/W2017587301","https://openalex.org/W2333876175","https://openalex.org/W3214851784","https://openalex.org/W2185880422","https://openalex.org/W2740001873","https://openalex.org/W1971132357"],"abstract_inverted_index":{"Traditionally,":[0],"MPI":[1],"runtimes":[2],"have":[3,75],"been":[4,76],"designed":[5],"for":[6,88,114,133,146,166,189,199],"clusters":[7],"with":[8,15,36,94,120],"a":[9,83,121],"large":[10,58,201],"number":[11],"of":[12,18,61,124,170],"nodes.":[13],"However,":[14],"the":[16,65,89,115,155,171,184],"advent":[17],"MPI+CUDA":[19],"applications":[20],"and":[21,48,127,139,148,191,195],"dense":[22],"multi-GPU":[23,106],"systems,":[24],"it":[25],"has":[26],"become":[27],"important":[28],"to":[29,56,137,143,160],"design":[30,53,87],"efficient":[31,104],"communication":[32,60],"schemes.":[33],"This":[34],"coupled":[35],"new":[37],"application":[38],"workloads":[39],"brought":[40],"forward":[41],"by":[42],"Deep":[43],"Learning":[44],"frameworks":[45],"like":[46,73],"Caffe":[47],"Microsoft":[49,178],"CNTK":[50],"pose":[51],"additional":[52],"constraints":[54],"due":[55],"very":[57,200],"message":[59,193,202],"GPU":[62],"buffers":[63],"during":[64],"training":[66,169],"phase.":[67],"In":[68,78,153],"this":[69,79],"context,":[70],"special-purpose":[71],"libraries":[72],"NCCL":[74,125],"proposed.":[77],"paper,":[80],"we":[81],"propose":[82],"pipelined":[84],"chain":[85],"(ring)":[86],"MPI_Bcast":[90,117],"collective":[91,97],"operation":[92],"along":[93,119],"an":[95,110],"enhanced":[96],"tuning":[98],"framework":[99],"in":[100],"MVAPICH2-GDR":[101,134],"that":[102],"enables":[103],"intra-/internode":[105],"communication.":[107],"We":[108],"present":[109],"in-depth":[111],"performance":[112,198],"landscape":[113],"proposed":[116,131,156,181],"schemes":[118],"comparative":[122],"analysis":[123],"Broadcast":[126],"NCCL-based":[128,144,164],"MPI_Bcast.":[129],"The":[130,180],"designs":[132,157],"enable":[135],"up":[136,159],"14X":[138],"16.6X":[140],"improvement,":[141],"compared":[142],"solutions,":[145],"intra-":[147],"internode":[149],"broadcast":[150],"latency,":[151],"respectively.":[152],"addition,":[154],"provide":[158],"7%":[161],"improvement":[162],"over":[163],"solutions":[165,182],"data":[167],"parallel":[168],"VGG":[172],"network":[173],"on":[174],"128":[175],"GPUs":[176],"using":[177],"CNTK.":[179],"outperform":[183],"recently":[185],"introduced":[186],"NCCL2":[187],"library":[188],"small":[190],"medium":[192],"sizes":[194],"offer":[196],"comparable/better":[197],"sizes.":[203]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":5},{"year":2024,"cited_by_count":4},{"year":2023,"cited_by_count":3},{"year":2022,"cited_by_count":2},{"year":2021,"cited_by_count":5},{"year":2020,"cited_by_count":2},{"year":2019,"cited_by_count":9},{"year":2018,"cited_by_count":4},{"year":2017,"cited_by_count":1}],"updated_date":"2026-03-06T13:50:29.536080","created_date":"2025-10-10T00:00:00"}