{"id":"https://openalex.org/W4413756535","doi":"https://doi.org/10.1145/3718958.3750503","title":"From ATOP to ZCube: Automated Topology Optimization Pipeline and A Highly Cost-Effective Network Topology for Large Model Training","display_name":"From ATOP to ZCube: Automated Topology Optimization Pipeline and A Highly Cost-Effective Network Topology for Large Model Training","publication_year":2025,"publication_date":"2025-08-27","ids":{"openalex":"https://openalex.org/W4413756535","doi":"https://doi.org/10.1145/3718958.3750503"},"language":"en","primary_location":{"id":"doi:10.1145/3718958.3750503","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3718958.3750503","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3718958.3750503","source":null,"license":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM SIGCOMM 2025 Conference","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3718958.3750503","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Zihan Yan","orcid":"https://orcid.org/0009-0000-8736-8876"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Zihan Yan","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0000-8736-8876","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100380761","display_name":"Dan Li","orcid":"https://orcid.org/0000-0002-7581-8865"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dan Li","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-7581-8865","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100379245","display_name":"Li Chen","orcid":"https://orcid.org/0000-0002-4228-7885"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li Chen","raw_affiliation_strings":["Zhongguancun Laboratory, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-4228-7885","affiliations":[{"raw_affiliation_string":"Zhongguancun Laboratory, Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088225159","display_name":"Dian Xiong","orcid":"https://orcid.org/0000-0002-8905-9046"},"institutions":[{"id":"https://openalex.org/I2802757744","display_name":"Transnet (South Africa)","ror":"https://ror.org/01nap1w14","country_code":"ZA","type":"company","lineage":["https://openalex.org/I2802757744"]}],"countries":["ZA"],"is_corresponding":false,"raw_author_name":"Dian Xiong","raw_affiliation_strings":["Harnets.AI, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-8905-9046","affiliations":[{"raw_affiliation_string":"Harnets.AI, Beijing, China","institution_ids":["https://openalex.org/I2802757744"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5054822731","display_name":"Kaihui Gao","orcid":"https://orcid.org/0000-0001-9013-5993"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kaihui Gao","raw_affiliation_strings":["Zhongguancun Laboratory, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-9013-5993","affiliations":[{"raw_affiliation_string":"Zhongguancun Laboratory, Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047276586","display_name":"Yiwei Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yiwei Zhang","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0005-2270-015X","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065324930","display_name":"Rui Yan","orcid":"https://orcid.org/0009-0007-8632-2017"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Rui Yan","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0007-8632-2017","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Menglei Zhang","orcid":"https://orcid.org/0009-0005-8124-8757"},"institutions":[{"id":"https://openalex.org/I58610484","display_name":"Seattle University","ror":"https://ror.org/02jqc0m91","country_code":"US","type":"education","lineage":["https://openalex.org/I58610484"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Menglei Zhang","raw_affiliation_strings":["ByteDance, Seattle, USA"],"raw_orcid":"https://orcid.org/0009-0005-8124-8757","affiliations":[{"raw_affiliation_string":"ByteDance, Seattle, USA","institution_ids":["https://openalex.org/I58610484"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088930668","display_name":"Bochun Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bochun Zhang","raw_affiliation_strings":["ByteDance, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0008-5421-3817","affiliations":[{"raw_affiliation_string":"ByteDance, Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101871699","display_name":"Zhuo Jiang","orcid":"https://orcid.org/0000-0002-6144-7899"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhuo Jiang","raw_affiliation_strings":["ByteDance, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-6144-7899","affiliations":[{"raw_affiliation_string":"ByteDance, Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049104908","display_name":"Jianxi Ye","orcid":null},"institutions":[{"id":"https://openalex.org/I58610484","display_name":"Seattle University","ror":"https://ror.org/02jqc0m91","country_code":"US","type":"education","lineage":["https://openalex.org/I58610484"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jianxi Ye","raw_affiliation_strings":["ByteDance, Seattle, USA"],"raw_orcid":"https://orcid.org/0009-0007-3395-3624","affiliations":[{"raw_affiliation_string":"ByteDance, Seattle, USA","institution_ids":["https://openalex.org/I58610484"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5084047386","display_name":"Haibin Lin","orcid":"https://orcid.org/0000-0003-4879-5335"},"institutions":[{"id":"https://openalex.org/I58610484","display_name":"Seattle University","ror":"https://ror.org/02jqc0m91","country_code":"US","type":"education","lineage":["https://openalex.org/I58610484"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Haibin Lin","raw_affiliation_strings":["ByteDance, Seattle, USA"],"raw_orcid":"https://orcid.org/0000-0003-4879-5335","affiliations":[{"raw_affiliation_string":"ByteDance, Seattle, USA","institution_ids":["https://openalex.org/I58610484"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":12,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I99065089"],"apc_list":null,"apc_paid":null,"fwci":1.3517,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.85083748,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":null,"issue":null,"first_page":"861","last_page":"881"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10715","display_name":"Distributed and Parallel Computing Systems","score":0.9883999824523926,"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"}},"topics":[{"id":"https://openalex.org/T10715","display_name":"Distributed and Parallel Computing Systems","score":0.9883999824523926,"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"}},{"id":"https://openalex.org/T11522","display_name":"VLSI and FPGA Design Techniques","score":0.9556999802589417,"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/T11159","display_name":"Manufacturing Process and Optimization","score":0.954800009727478,"subfield":{"id":"https://openalex.org/subfields/2209","display_name":"Industrial and Manufacturing Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.7388855218887329},{"id":"https://openalex.org/keywords/topology-optimization","display_name":"Topology optimization","score":0.7012059688568115},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.678234338760376},{"id":"https://openalex.org/keywords/topology","display_name":"Topology (electrical circuits)","score":0.654850959777832},{"id":"https://openalex.org/keywords/network-topology","display_name":"Network topology","score":0.6138174533843994},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.6048920154571533},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.25720950961112976},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.1799549162387848},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.07448014616966248},{"id":"https://openalex.org/keywords/finite-element-method","display_name":"Finite element method","score":0.07287269830703735},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.06575345993041992}],"concepts":[{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.7388855218887329},{"id":"https://openalex.org/C189216461","wikidata":"https://www.wikidata.org/wiki/Q2443456","display_name":"Topology optimization","level":3,"score":0.7012059688568115},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.678234338760376},{"id":"https://openalex.org/C184720557","wikidata":"https://www.wikidata.org/wiki/Q7825049","display_name":"Topology (electrical circuits)","level":2,"score":0.654850959777832},{"id":"https://openalex.org/C199845137","wikidata":"https://www.wikidata.org/wiki/Q145490","display_name":"Network topology","level":2,"score":0.6138174533843994},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.6048920154571533},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.25720950961112976},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.1799549162387848},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.07448014616966248},{"id":"https://openalex.org/C135628077","wikidata":"https://www.wikidata.org/wiki/Q220184","display_name":"Finite element method","level":2,"score":0.07287269830703735},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.06575345993041992},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","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}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3718958.3750503","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3718958.3750503","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3718958.3750503","source":null,"license":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM SIGCOMM 2025 Conference","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3718958.3750503","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3718958.3750503","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3718958.3750503","source":null,"license":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM SIGCOMM 2025 Conference","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4413756535.pdf","grobid_xml":"https://content.openalex.org/works/W4413756535.grobid-xml"},"referenced_works_count":30,"referenced_works":["https://openalex.org/W95608104","https://openalex.org/W1558919105","https://openalex.org/W1923594657","https://openalex.org/W1966570738","https://openalex.org/W2007089612","https://openalex.org/W2010788560","https://openalex.org/W2038423000","https://openalex.org/W2096313289","https://openalex.org/W2096811399","https://openalex.org/W2119541875","https://openalex.org/W2126105956","https://openalex.org/W2126210439","https://openalex.org/W2131613942","https://openalex.org/W2132320636","https://openalex.org/W2155187164","https://openalex.org/W2169246522","https://openalex.org/W2308071406","https://openalex.org/W2995623003","https://openalex.org/W3204998121","https://openalex.org/W3216755252","https://openalex.org/W4256524787","https://openalex.org/W4376652719","https://openalex.org/W4380874786","https://openalex.org/W4383550741","https://openalex.org/W4386396242","https://openalex.org/W4386396793","https://openalex.org/W4386443196","https://openalex.org/W4401176521","https://openalex.org/W4401176799","https://openalex.org/W4402389838"],"related_works":["https://openalex.org/W2046491619","https://openalex.org/W230091440","https://openalex.org/W2389142552","https://openalex.org/W2239929624","https://openalex.org/W2883804989","https://openalex.org/W1974803732","https://openalex.org/W2081304691","https://openalex.org/W2127556255","https://openalex.org/W3018492858","https://openalex.org/W279903623"],"abstract_inverted_index":{"The":[0],"development":[1],"of":[2,37,42,61,63,65],"large":[3],"language":[4],"models":[5,31],"(LLMs)":[6],"poses":[7],"new":[8,115],"challenges":[9],"in":[10,18,99],"data":[11,107],"center":[12],"network":[13,32,71,93,157],"topology":[14,20,33,56,116],"design.":[15],"To":[16],"assist":[17],"exploring":[19],"design,":[21],"we":[22,112],"propose":[23],"ATOP,":[24],"an":[25],"Automated":[26],"Topology":[27,184],"Optimization":[28],"Pipeline,":[29],"which":[30,119],"as":[34],"a":[35,59,106,114,169],"set":[36],"hyperparameters,":[38],"enabling":[39],"the":[40,121,135,187],"discovery":[41],"potential":[43],"topologies.":[44],"With":[45],"various":[46,125],"optimization":[47,51,57],"algorithms":[48],"and":[49,77,92,104,144,155,190],"customizable":[50],"objectives,":[52],"ATOP":[53,69,98],"achieves":[54],"automated":[55],"on":[58,70,168],"scale":[60],"tens":[62],"thousands":[64],"GPUs.":[66],"We":[67,95,164],"apply":[68],"topologies":[72],"for":[73],"256,":[74],"1024,":[75],"4096,":[76],"16384":[78],"GPUs,":[79],"optimizing":[80],"performance":[81],"under":[82],"LLMs":[83],"training":[84,149],"traffic":[85],"patterns,":[86],"collective":[87],"communication":[88],"performance,":[89],"fault":[90],"tolerance,":[91],"cost.":[94],"also":[96,165],"evaluate":[97],"different":[100],"scenarios:":[101],"building,":[102],"optimizing,":[103],"expanding":[105],"center.":[108],"From":[109],"ATOP's":[110],"results,":[111],"discover":[113],"\u2014":[117],"ZCube,":[118,132],"reaches":[120],"highest":[122],"cost-effectiveness":[123],"across":[124],"GPU":[126],"scales.":[127],"Simulation":[128],"results":[129],"show":[130,173],"that":[131,174],"compared":[133,181],"to":[134,153,162,182],"previous":[136],"state-of-the-art":[137],"topologies,":[138],"including":[139],"Rail-optimized":[140],"Fat-tree":[141],"(ROFT),":[142],"Rail-only,":[143],"HPN,":[145],"improves":[146],"end-to-end":[147],"LLM":[148],"speed":[150],"by":[151,160,179],"3%":[152],"7%":[154],"reduces":[156,176],"hardware":[158,177],"costs":[159,178],"26%":[161],"46%.":[163],"construct":[166],"ZCube":[167,175],"real-world":[170],"testbed.":[171],"Results":[172],"25%":[180],"Rail-Optimized":[183],"while":[185],"maintaining":[186],"same":[188],"all-reduce":[189],"all-to-all":[191],"performance.":[192]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-03-27T05:58:40.876381","created_date":"2025-10-10T00:00:00"}