{"id":"https://openalex.org/W2999276551","doi":"https://doi.org/10.1109/mlhpc49564.2019.00006","title":"Understanding Scalability and Fine-Grain Parallelism of Synchronous Data Parallel Training","display_name":"Understanding Scalability and Fine-Grain Parallelism of Synchronous Data Parallel Training","publication_year":2019,"publication_date":"2019-11-01","ids":{"openalex":"https://openalex.org/W2999276551","doi":"https://doi.org/10.1109/mlhpc49564.2019.00006","mag":"2999276551"},"language":"en","primary_location":{"id":"doi:10.1109/mlhpc49564.2019.00006","is_oa":false,"landing_page_url":"https://doi.org/10.1109/mlhpc49564.2019.00006","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2019 IEEE/ACM Workshop on Machine Learning in High Performance Computing Environments (MLHPC)","raw_type":"proceedings-article"},"type":"preprint","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://hal.science/hal-02570148","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5023743368","display_name":"Jiali Li","orcid":"https://orcid.org/0000-0001-7929-8484"},"institutions":[{"id":"https://openalex.org/I75027704","display_name":"University of Tennessee at Knoxville","ror":"https://ror.org/020f3ap87","country_code":"US","type":"education","lineage":["https://openalex.org/I75027704"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Jiali Li","raw_affiliation_strings":["University of Tennessee, USA","ICL - Innovative Computing Laboratory [Knoxville] (Suite 203 Claxton\r\n1122 Volunteer Blvd\r\nKnoxville, TN 37996 - \u00c9tats-Unis)"],"affiliations":[{"raw_affiliation_string":"University of Tennessee, USA","institution_ids":["https://openalex.org/I75027704"]},{"raw_affiliation_string":"ICL - Innovative Computing Laboratory [Knoxville] (Suite 203 Claxton\r\n1122 Volunteer Blvd\r\nKnoxville, TN 37996 - \u00c9tats-Unis)","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5085745891","display_name":"Bogdan Nicolae","orcid":"https://orcid.org/0000-0002-0661-7509"},"institutions":[{"id":"https://openalex.org/I1282105669","display_name":"Argonne National Laboratory","ror":"https://ror.org/05gvnxz63","country_code":"US","type":"facility","lineage":["https://openalex.org/I1282105669","https://openalex.org/I1330989302","https://openalex.org/I39565521","https://openalex.org/I40347166"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Bogdan Nicolae","raw_affiliation_strings":["Argonne National Laboratory, USA","ANL - Argonne National Laboratory [Lemont] (9700 S Cass Ave B109, Lemont, IL, 60439 - \u00c9tats-Unis)"],"affiliations":[{"raw_affiliation_string":"Argonne National Laboratory, USA","institution_ids":["https://openalex.org/I1282105669"]},{"raw_affiliation_string":"ANL - Argonne National Laboratory [Lemont] (9700 S Cass Ave B109, Lemont, IL, 60439 - \u00c9tats-Unis)","institution_ids":["https://openalex.org/I1282105669"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5020776475","display_name":"Justin M. Wozniak","orcid":"https://orcid.org/0000-0002-2441-2048"},"institutions":[{"id":"https://openalex.org/I1282105669","display_name":"Argonne National Laboratory","ror":"https://ror.org/05gvnxz63","country_code":"US","type":"facility","lineage":["https://openalex.org/I1282105669","https://openalex.org/I1330989302","https://openalex.org/I39565521","https://openalex.org/I40347166"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Justin Wozniak","raw_affiliation_strings":["Argonne National Laboratory, USA","ANL - Argonne National Laboratory [Lemont] (9700 S Cass Ave B109, Lemont, IL, 60439 - \u00c9tats-Unis)"],"affiliations":[{"raw_affiliation_string":"Argonne National Laboratory, USA","institution_ids":["https://openalex.org/I1282105669"]},{"raw_affiliation_string":"ANL - Argonne National Laboratory [Lemont] (9700 S Cass Ave B109, Lemont, IL, 60439 - \u00c9tats-Unis)","institution_ids":["https://openalex.org/I1282105669"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5010055736","display_name":"George Bosilca","orcid":"https://orcid.org/0000-0003-2411-8495"},"institutions":[{"id":"https://openalex.org/I75027704","display_name":"University of Tennessee at Knoxville","ror":"https://ror.org/020f3ap87","country_code":"US","type":"education","lineage":["https://openalex.org/I75027704"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"George Bosilca","raw_affiliation_strings":["University of Tennessee, USA","ICL - Innovative Computing Laboratory [Knoxville] (Suite 203 Claxton\r\n1122 Volunteer Blvd\r\nKnoxville, TN 37996 - \u00c9tats-Unis)"],"affiliations":[{"raw_affiliation_string":"University of Tennessee, USA","institution_ids":["https://openalex.org/I75027704"]},{"raw_affiliation_string":"ICL - Innovative Computing Laboratory [Knoxville] (Suite 203 Claxton\r\n1122 Volunteer Blvd\r\nKnoxville, TN 37996 - \u00c9tats-Unis)","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5023743368"],"corresponding_institution_ids":["https://openalex.org/I75027704"],"apc_list":null,"apc_paid":null,"fwci":0.8671,"has_fulltext":false,"cited_by_count":6,"citation_normalized_percentile":{"value":0.81749094,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":89,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"8"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9991000294685364,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9991000294685364,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"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":0.9988999962806702,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.9986000061035156,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8702436089515686},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.8167715072631836},{"id":"https://openalex.org/keywords/granularity","display_name":"Granularity","score":0.7388549447059631},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.6420108675956726},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.6348340511322021},{"id":"https://openalex.org/keywords/big-data","display_name":"Big data","score":0.5911201238632202},{"id":"https://openalex.org/keywords/interleaving","display_name":"Interleaving","score":0.5680667161941528},{"id":"https://openalex.org/keywords/synchronization","display_name":"Synchronization (alternating current)","score":0.5668134689331055},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5250694155693054},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.47771215438842773},{"id":"https://openalex.org/keywords/supercomputer","display_name":"Supercomputer","score":0.46854981780052185},{"id":"https://openalex.org/keywords/node","display_name":"Node (physics)","score":0.45850449800491333},{"id":"https://openalex.org/keywords/distributed-computing","display_name":"Distributed computing","score":0.4560210704803467},{"id":"https://openalex.org/keywords/massively-parallel","display_name":"Massively parallel","score":0.4321483373641968},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.1577644944190979},{"id":"https://openalex.org/keywords/data-mining","display_name":"Data mining","score":0.1549583077430725},{"id":"https://openalex.org/keywords/database","display_name":"Database","score":0.15040802955627441},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.11425217986106873}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8702436089515686},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.8167715072631836},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.7388549447059631},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.6420108675956726},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6348340511322021},{"id":"https://openalex.org/C75684735","wikidata":"https://www.wikidata.org/wiki/Q858810","display_name":"Big data","level":2,"score":0.5911201238632202},{"id":"https://openalex.org/C28034677","wikidata":"https://www.wikidata.org/wiki/Q17092530","display_name":"Interleaving","level":2,"score":0.5680667161941528},{"id":"https://openalex.org/C2778562939","wikidata":"https://www.wikidata.org/wiki/Q1298791","display_name":"Synchronization (alternating current)","level":3,"score":0.5668134689331055},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5250694155693054},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.47771215438842773},{"id":"https://openalex.org/C83283714","wikidata":"https://www.wikidata.org/wiki/Q121117","display_name":"Supercomputer","level":2,"score":0.46854981780052185},{"id":"https://openalex.org/C62611344","wikidata":"https://www.wikidata.org/wiki/Q1062658","display_name":"Node (physics)","level":2,"score":0.45850449800491333},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.4560210704803467},{"id":"https://openalex.org/C190475519","wikidata":"https://www.wikidata.org/wiki/Q544384","display_name":"Massively parallel","level":2,"score":0.4321483373641968},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.1577644944190979},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.1549583077430725},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.15040802955627441},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.11425217986106873},{"id":"https://openalex.org/C127162648","wikidata":"https://www.wikidata.org/wiki/Q16858953","display_name":"Channel (broadcasting)","level":2,"score":0.0},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.0},{"id":"https://openalex.org/C66938386","wikidata":"https://www.wikidata.org/wiki/Q633538","display_name":"Structural engineering","level":1,"score":0.0},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/mlhpc49564.2019.00006","is_oa":false,"landing_page_url":"https://doi.org/10.1109/mlhpc49564.2019.00006","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2019 IEEE/ACM Workshop on Machine Learning in High Performance Computing Environments (MLHPC)","raw_type":"proceedings-article"},{"id":"pmh:oai:HAL:hal-02570148v1","is_oa":true,"landing_page_url":"https://hal.science/hal-02570148","pdf_url":null,"source":{"id":"https://openalex.org/S4306402512","display_name":"HAL (Le Centre pour la Communication Scientifique Directe)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1294671590","host_organization_name":"Centre National de la Recherche Scientifique","host_organization_lineage":["https://openalex.org/I1294671590"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2019 IEEE/ACM Workshop on Machine Learning in High Performance Computing Environments (MLHPC), Nov 2019, Denver, United States. pp.1-8, ⟨10.1109/MLHPC49564.2019.00006⟩","raw_type":"Conference papers"}],"best_oa_location":{"id":"pmh:oai:HAL:hal-02570148v1","is_oa":true,"landing_page_url":"https://hal.science/hal-02570148","pdf_url":null,"source":{"id":"https://openalex.org/S4306402512","display_name":"HAL (Le Centre pour la Communication Scientifique Directe)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1294671590","host_organization_name":"Centre National de la Recherche Scientifique","host_organization_lineage":["https://openalex.org/I1294671590"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2019 IEEE/ACM Workshop on Machine Learning in High Performance Computing Environments (MLHPC), Nov 2019, Denver, United States. pp.1-8, ⟨10.1109/MLHPC49564.2019.00006⟩","raw_type":"Conference papers"},"sustainable_development_goals":[{"score":0.6100000143051147,"display_name":"Industry, innovation and infrastructure","id":"https://metadata.un.org/sdg/9"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":26,"referenced_works":["https://openalex.org/W1533411515","https://openalex.org/W1861703862","https://openalex.org/W2108598243","https://openalex.org/W2127941149","https://openalex.org/W2155893237","https://openalex.org/W2181607856","https://openalex.org/W2194775991","https://openalex.org/W2576960802","https://openalex.org/W2805253785","https://openalex.org/W2898319404","https://openalex.org/W2906137127","https://openalex.org/W2951947775","https://openalex.org/W2960293275","https://openalex.org/W2961595369","https://openalex.org/W2962747323","https://openalex.org/W2962863496","https://openalex.org/W2962911728","https://openalex.org/W2963304552","https://openalex.org/W2963382388","https://openalex.org/W2963807318","https://openalex.org/W2963933775","https://openalex.org/W3102254513","https://openalex.org/W3103417206","https://openalex.org/W6679154944","https://openalex.org/W6686045668","https://openalex.org/W6745458143"],"related_works":["https://openalex.org/W1655266410","https://openalex.org/W2389051085","https://openalex.org/W2330343234","https://openalex.org/W1901012776","https://openalex.org/W2463883322","https://openalex.org/W2931688134","https://openalex.org/W2377919138","https://openalex.org/W2229382548","https://openalex.org/W1632118563","https://openalex.org/W2073684863"],"abstract_inverted_index":{"In":[0],"the":[1,50,65,86,174,197,207],"age":[2],"of":[3,30,35,67,88,128,131,158,166,176,199],"big":[4],"data,":[5,37],"deep":[6],"learning":[7,31],"has":[8,155],"emerged":[9],"as":[10,126],"a":[11,90,100],"powerful":[12],"tool":[13],"to":[14,60,195],"extract":[15],"insight":[16],"and":[17,24,33,118,150,170],"exploit":[18],"its":[19],"value,":[20],"both":[21],"in":[22,81,156],"industry":[23],"scientific":[25],"applications.":[26],"With":[27],"increasing":[28],"complexity":[29],"models":[32],"amounts":[34],"training":[36],"data-parallel":[38,68,92],"approaches":[39,69],"based":[40],"on":[41,70,96,98],"frequent":[42],"all-reduce":[43,179],"synchronization":[44],"steps":[45],"are":[46,142,204],"increasingly":[47],"popular.":[48],"Despite":[49],"fact":[51],"that":[52,94,164,203],"high-performance":[53],"computing":[54],"(HPC)":[55],"technologies":[56],"have":[57],"been":[58],"designed":[59],"address":[61,78],"such":[62,125],"patterns":[63],"efficiently,":[64],"behavior":[66,87],"HPC":[71],"platforms":[72],"is":[73,120],"not":[74],"well":[75],"understood.":[76],"To":[77],"this":[79,82,153],"issue,":[80],"paper":[83],"we":[84,111],"study":[85],"Horovod,":[89],"popular":[91],"approach":[93],"relies":[95],"MPI,":[97],"Theta,":[99],"pre-Exascale":[101],"machine":[102],"at":[103,147],"Argonne":[104],"National":[105],"Laboratory.":[106],"Using":[107],"two":[108,113],"representative":[109],"applications,":[110],"explore":[112],"aspects:":[114],"(1)":[115],"how":[116,139],"performance":[117],"scalability":[119],"affected":[121],"by":[122],"important":[123],"parameters":[124],"number":[127,130],"nodes,":[129],"workers,":[132],"threads":[133],"per":[134],"node,":[135],"batch":[136],"size;":[137],"(2)":[138],"computational":[140],"phases":[141,146],"interleaved":[143],"withall-reduce":[144],"communication":[145],"fine":[148],"granularity":[149],"what":[151],"consequences":[152],"interleaving":[154],"terms":[157],"potential":[159],"bottlenecks.":[160],"Our":[161],"findings":[162],"show":[163],"pipelining":[165],"back-propagation,":[167],"gradient":[168],"reduction":[169],"weight":[171],"updates":[172],"mitigate":[173],"effects":[175],"stragglers":[177],"during":[178],"only":[180],"partially.":[181],"Furthermore,":[182],"there":[183],"can":[184,192],"be":[185,193],"significant":[186],"delays":[187],"between":[188],"weights":[189],"update,":[190],"which":[191],"leveraged":[194],"mask":[196],"overhead":[198],"additional":[200],"background":[201],"operations":[202],"coupled":[205],"with":[206],"training.":[208]},"counts_by_year":[{"year":2022,"cited_by_count":1},{"year":2021,"cited_by_count":1},{"year":2020,"cited_by_count":4}],"updated_date":"2026-03-10T16:38:18.471706","created_date":"2025-10-10T00:00:00"}