{"id":"https://openalex.org/W2986124642","doi":"https://doi.org/10.1109/cluster.2019.8891042","title":"Performance Characterization of DNN Training using TensorFlow and PyTorch on Modern Clusters","display_name":"Performance Characterization of DNN Training using TensorFlow and PyTorch on Modern Clusters","publication_year":2019,"publication_date":"2019-09-01","ids":{"openalex":"https://openalex.org/W2986124642","doi":"https://doi.org/10.1109/cluster.2019.8891042","mag":"2986124642"},"language":"en","primary_location":{"id":"doi:10.1109/cluster.2019.8891042","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cluster.2019.8891042","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2019 IEEE International Conference on Cluster Computing (CLUSTER)","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/A5057769693","display_name":"Arpan Jain","orcid":"https://orcid.org/0000-0003-2522-8522"},"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":"Arpan Jain","raw_affiliation_strings":["Dept. of Computer Science and Engineering, The Ohio State University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Dept. of Computer Science and Engineering, The Ohio State University","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"middle","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":false,"raw_author_name":"Ammar Ahmad Awan","raw_affiliation_strings":["Dept. of Computer Science and Engineering, The Ohio State University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Dept. of Computer Science and Engineering, The Ohio State University","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5015153122","display_name":"Quentin Anthony","orcid":"https://orcid.org/0000-0002-6823-9080"},"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":"Quentin Anthony","raw_affiliation_strings":["Dept. of Computer Science and Engineering, The Ohio State University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Dept. of Computer Science and Engineering, The Ohio State University","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":["Dept. of Computer Science and Engineering, The Ohio State University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Dept. of Computer Science and Engineering, The Ohio State University","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5105345016","display_name":"Dhableswar K. Panda","orcid":null},"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":"Dhableswar K. DK Panda","raw_affiliation_strings":["Dept. of Computer Science and Engineering, The Ohio State University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Dept. of Computer Science and Engineering, The Ohio State University","institution_ids":["https://openalex.org/I52357470"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I52357470"],"apc_list":null,"apc_paid":null,"fwci":1.6488,"has_fulltext":false,"cited_by_count":36,"citation_normalized_percentile":{"value":0.90187268,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":89,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"11"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":1.0,"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":1.0,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9977999925613403,"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/T12676","display_name":"Machine Learning and ELM","score":0.9977999925613403,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8602526187896729},{"id":"https://openalex.org/keywords/xeon","display_name":"Xeon","score":0.7843436002731323},{"id":"https://openalex.org/keywords/xeon-phi","display_name":"Xeon Phi","score":0.6545048952102661},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.6313954591751099},{"id":"https://openalex.org/keywords/cuda","display_name":"CUDA","score":0.5000169277191162},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.45049113035202026},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.44229018688201904},{"id":"https://openalex.org/keywords/profiling","display_name":"Profiling (computer programming)","score":0.4382740259170532},{"id":"https://openalex.org/keywords/multi-core-processor","display_name":"Multi-core processor","score":0.433197557926178},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4280494153499603},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.355907142162323},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.21527951955795288}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8602526187896729},{"id":"https://openalex.org/C145108525","wikidata":"https://www.wikidata.org/wiki/Q656154","display_name":"Xeon","level":2,"score":0.7843436002731323},{"id":"https://openalex.org/C96972482","wikidata":"https://www.wikidata.org/wiki/Q1049168","display_name":"Xeon Phi","level":2,"score":0.6545048952102661},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6313954591751099},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.5000169277191162},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.45049113035202026},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.44229018688201904},{"id":"https://openalex.org/C187191949","wikidata":"https://www.wikidata.org/wiki/Q1138496","display_name":"Profiling (computer programming)","level":2,"score":0.4382740259170532},{"id":"https://openalex.org/C78766204","wikidata":"https://www.wikidata.org/wiki/Q555032","display_name":"Multi-core processor","level":2,"score":0.433197557926178},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4280494153499603},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.355907142162323},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.21527951955795288},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/cluster.2019.8891042","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cluster.2019.8891042","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2019 IEEE International Conference on Cluster Computing (CLUSTER)","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":0.44999998807907104}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":29,"referenced_works":["https://openalex.org/W1817724761","https://openalex.org/W1973445088","https://openalex.org/W2099866409","https://openalex.org/W2105605904","https://openalex.org/W2156876426","https://openalex.org/W2163605009","https://openalex.org/W2168231600","https://openalex.org/W2622263826","https://openalex.org/W2731415668","https://openalex.org/W2766362889","https://openalex.org/W2787998955","https://openalex.org/W2898319404","https://openalex.org/W2899771611","https://openalex.org/W2912208425","https://openalex.org/W2915943718","https://openalex.org/W2953384591","https://openalex.org/W2963674387","https://openalex.org/W2963964896","https://openalex.org/W4301239768","https://openalex.org/W6683161558","https://openalex.org/W6684191040","https://openalex.org/W6684859321","https://openalex.org/W6713134421","https://openalex.org/W6739622702","https://openalex.org/W6748645090","https://openalex.org/W6756040250","https://openalex.org/W6756193283","https://openalex.org/W6759814442","https://openalex.org/W6780043775"],"related_works":["https://openalex.org/W2475524688","https://openalex.org/W2739740241","https://openalex.org/W2085105049","https://openalex.org/W2592417500","https://openalex.org/W1974923383","https://openalex.org/W2526069705","https://openalex.org/W2024016913","https://openalex.org/W2019153376","https://openalex.org/W2796552083","https://openalex.org/W2022666014"],"abstract_inverted_index":{"The":[0,202],"recent":[1],"surge":[2],"of":[3,20,76,94,138,151],"Deep":[4,49],"Learning":[5],"(DL)":[6],"models":[7],"and":[8,23,30,82,101,107,141,158,161,167,177,194,197,241],"applications":[9],"can":[10],"be":[11,119],"attributed":[12],"to":[13,208,211,220,236],"the":[14,95,126,134,149,153],"rise":[15],"in":[16],"computational":[17],"resources,":[18],"availability":[19],"large-scale":[21],"datasets,":[22],"accessible":[24],"DL":[25],"frameworks":[26,34],"such":[27,79,147,180],"as":[28,80,148,181],"TensorFlow":[29],"PyTorch.":[31],"Because":[32],"these":[33],"have":[35],"been":[36],"heavily":[37],"optimized":[38],"for":[39,47,122,192,200],"NVIDIA":[40,102],"GPUs,":[41],"several":[42],"performance":[43,74],"characterization":[44,75],"studies":[45,60],"exist":[46,56],"GPU-based":[48,195],"Neural":[50],"Network":[51],"(DNN)":[52],"training.":[53,66],"However,":[54],"there":[55],"very":[57],"few":[58],"research":[59],"that":[61,232],"focus":[62],"on":[63,84,144,245],"CPU-based":[64],"DNN":[65,159,178],"In":[67],"this":[68],"paper,":[69],"we":[70,205,230],"provide":[71,110,188],"an":[72],"in-depth":[73],"state-of-the-art":[77],"DNNs":[78],"ResNet(s)":[81],"Inception-v3/v4":[83],"multiple":[85],"CPU":[86,193],"architectures":[87],"including":[88],"Intel":[89,96],"Xeon":[90,97],"Broadwell,":[91],"three":[92,111],"variants":[93],"Skylake,":[98],"AMD":[99],"EPYC,":[100],"GPUs":[103],"like":[104],"K80,":[105],"P100,":[106],"V100.":[108],"We":[109,186],"key":[112],"insights:":[113],"1)":[114],"Multi-process":[115],"(MP)":[116],"training":[117,196],"should":[118],"used":[120],"even":[121],"a":[123,165,189,215,224],"single-node,":[124],"because":[125],"single-process":[127],"(SP)":[128],"approach":[129],"cannot":[130],"fully":[131],"exploit":[132],"all":[133],"cores,":[135,152],"2)":[136],"Performance":[137],"both":[139],"SP":[140,240],"MP":[142,233],"depend":[143],"various":[145],"features":[146],"number":[150],"processes":[154],"per":[155],"node":[156],"(ppn),":[157],"architecture,":[160],"3)":[162],"There":[163],"is":[164,209,234],"non-linear":[166],"complex":[168],"relationship":[169],"between":[170,184],"CPU/system":[171],"characteristics":[172],"(core-count,":[173],"ppn,":[174],"hyper-threading,":[175],"etc)":[176],"specifications":[179],"inherent":[182],"parallelism":[183],"layers.":[185],"further":[187],"comparative":[190],"analysis":[191,199],"profiling":[198],"Horovod.":[201],"fastest":[203],"Skylake":[204,247],"had":[206],"access":[207],"up":[210,219,235],"2.35\u00d7":[212],"better":[213],"than":[214,223,239],"K80":[216],"GPU":[217],"but":[218],"3.32\u00d7":[221],"slower":[222],"V100":[225],"GPU.":[226],"For":[227],"ResNet-152":[228],"training,":[229],"observed":[231],"1.47\u00d7":[237],"faster":[238],"achieves":[242],"125\u00d7":[243],"speedup":[244],"128":[246],"nodes.":[248]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":3},{"year":2024,"cited_by_count":7},{"year":2023,"cited_by_count":6},{"year":2022,"cited_by_count":5},{"year":2021,"cited_by_count":6},{"year":2020,"cited_by_count":6},{"year":2019,"cited_by_count":1}],"updated_date":"2026-08-29T07:29:34.045763","created_date":"2025-10-10T00:00:00"}