{"id":"https://openalex.org/W7139941507","doi":"https://doi.org/10.1109/lca.2026.3676331","title":"JBOC: Just a Bunch of CXL-Enabled SSDs for Resource-Efficient LLM Checkpointing","display_name":"JBOC: Just a Bunch of CXL-Enabled SSDs for Resource-Efficient LLM Checkpointing","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7139941507","doi":"https://doi.org/10.1109/lca.2026.3676331"},"language":null,"primary_location":{"id":"doi:10.1109/lca.2026.3676331","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2026.3676331","pdf_url":null,"source":{"id":"https://openalex.org/S17643076","display_name":"IEEE Computer Architecture Letters","issn_l":"1556-6056","issn":["1556-6056","1556-6064","2473-2575"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Computer Architecture Letters","raw_type":"journal-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/A5063686935","display_name":"Hakbeom Jang","orcid":null},"institutions":[{"id":"https://openalex.org/I10654025","display_name":"SK Group (United States)","ror":"https://ror.org/00qajw440","country_code":"US","type":"company","lineage":["https://openalex.org/I10654025","https://openalex.org/I134353371"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hakbeom Jang","raw_affiliation_strings":["SK hynix, San Jose, CA, USA"],"raw_orcid":"https://orcid.org/0009-0003-7036-7879","affiliations":[{"raw_affiliation_string":"SK hynix, San Jose, CA, USA","institution_ids":["https://openalex.org/I10654025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130243916","display_name":"Younghoon Min","orcid":null},"institutions":[{"id":"https://openalex.org/I10654025","display_name":"SK Group (United States)","ror":"https://ror.org/00qajw440","country_code":"US","type":"company","lineage":["https://openalex.org/I10654025","https://openalex.org/I134353371"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Younghoon Min","raw_affiliation_strings":["SK hynix, San Jose, CA, USA"],"raw_orcid":"https://orcid.org/0009-0009-7764-1766","affiliations":[{"raw_affiliation_string":"SK hynix, San Jose, CA, USA","institution_ids":["https://openalex.org/I10654025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130251116","display_name":"Dongha Yoon","orcid":null},"institutions":[{"id":"https://openalex.org/I10654025","display_name":"SK Group (United States)","ror":"https://ror.org/00qajw440","country_code":"US","type":"company","lineage":["https://openalex.org/I10654025","https://openalex.org/I134353371"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Dongha Yoon","raw_affiliation_strings":["SK hynix, San Jose, CA, USA"],"raw_orcid":"https://orcid.org/0009-0005-6284-0720","affiliations":[{"raw_affiliation_string":"SK hynix, San Jose, CA, USA","institution_ids":["https://openalex.org/I10654025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130234985","display_name":"Hane Yie","orcid":null},"institutions":[{"id":"https://openalex.org/I10654025","display_name":"SK Group (United States)","ror":"https://ror.org/00qajw440","country_code":"US","type":"company","lineage":["https://openalex.org/I10654025","https://openalex.org/I134353371"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hane Yie","raw_affiliation_strings":["SK hynix, San Jose, CA, USA"],"raw_orcid":"https://orcid.org/0009-0004-9875-1589","affiliations":[{"raw_affiliation_string":"SK hynix, San Jose, CA, USA","institution_ids":["https://openalex.org/I10654025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008944433","display_name":"Hoshik Kim","orcid":"https://orcid.org/0000-0002-4017-8124"},"institutions":[{"id":"https://openalex.org/I10654025","display_name":"SK Group (United States)","ror":"https://ror.org/00qajw440","country_code":"US","type":"company","lineage":["https://openalex.org/I10654025","https://openalex.org/I134353371"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hoshik Kim","raw_affiliation_strings":["SK hynix, San Jose, CA, USA"],"raw_orcid":"https://orcid.org/0000-0002-4017-8124","affiliations":[{"raw_affiliation_string":"SK hynix, San Jose, CA, USA","institution_ids":["https://openalex.org/I10654025"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5018146250","display_name":"Jongryool Kim","orcid":"https://orcid.org/0009-0006-5938-6878"},"institutions":[{"id":"https://openalex.org/I10654025","display_name":"SK Group (United States)","ror":"https://ror.org/00qajw440","country_code":"US","type":"company","lineage":["https://openalex.org/I10654025","https://openalex.org/I134353371"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jongryool Kim","raw_affiliation_strings":["SK hynix, San Jose, CA, USA"],"raw_orcid":"https://orcid.org/0009-0006-5938-6878","affiliations":[{"raw_affiliation_string":"SK hynix, San Jose, CA, USA","institution_ids":["https://openalex.org/I10654025"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.39853196,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"25","issue":"1","first_page":"182","last_page":"185"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11181","display_name":"Advanced Data Storage Technologies","score":0.9703999757766724,"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/T11181","display_name":"Advanced Data Storage Technologies","score":0.9703999757766724,"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/T11269","display_name":"Algorithms and Data Compression","score":0.0044999998062849045,"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/T10772","display_name":"Distributed systems and fault tolerance","score":0.004100000020116568,"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/key","display_name":"Key (lock)","score":0.3212999999523163},{"id":"https://openalex.org/keywords/processor-scheduling","display_name":"Processor scheduling","score":0.2752000093460083},{"id":"https://openalex.org/keywords/random-access-memory","display_name":"Random access memory","score":0.2524000108242035},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.23399999737739563}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7709000110626221},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.4011000096797943},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.3993000090122223},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3212999999523163},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.29260000586509705},{"id":"https://openalex.org/C2984822820","wikidata":"https://www.wikidata.org/wiki/Q1123036","display_name":"Processor scheduling","level":3,"score":0.2752000093460083},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.25999999046325684},{"id":"https://openalex.org/C2994168587","wikidata":"https://www.wikidata.org/wiki/Q5295","display_name":"Random access memory","level":2,"score":0.2524000108242035},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.23399999737739563},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.23029999434947968}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/lca.2026.3676331","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2026.3676331","pdf_url":null,"source":{"id":"https://openalex.org/S17643076","display_name":"IEEE Computer Architecture Letters","issn_l":"1556-6056","issn":["1556-6056","1556-6064","2473-2575"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Computer Architecture Letters","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"rapid":[1],"growth":[2],"of":[3],"Large":[4],"Language":[5],"Models":[6],"(LLMs)":[7],"has":[8],"pushed":[9],"training":[10,32,122],"systems":[11],"to":[12,54,124,159,170],"rely":[13],"on":[14,146],"memory":[15],"sharding":[16],"and":[17,38,82,95,134,152,189],"offloading,":[18],"introducing":[19],"substantial":[20],"data":[21,58],"movement":[22,59],"overhead.":[23],"This":[24,118],"is":[25,112],"further":[26],"exacerbated":[27],"by":[28],"frequent":[29],"checkpointing,":[30],"where":[31],"states":[33,123],"must":[34],"traverse":[35],"multiple":[36],"software":[37],"hardware":[39],"layers.":[40],"Such":[41],"a":[42,45,85,160],"process":[43],"creates":[44],"critical":[46],"performance":[47],"bottleneck,":[48],"as":[49,110],"traditional":[50],"host-centric":[51],"hierarchies":[52],"struggle":[53],"handle":[55],"the":[56,115,130,137],"massive":[57,91],"required":[60],"for":[61],"high-frequency,":[62],"faulttolerant":[63],"checkpointing.":[64,168],"In":[65],"this":[66],"paper,":[67],"we":[68],"introduce":[69],"JBOC,":[70],"an":[71],"in-device":[72],"processing":[73],"architecture":[74],"leveraging":[75],"CXL-enabled":[76],"SSDs":[77],"that":[78],"fuses":[79],"optimizer":[80,93],"updates":[81,94],"checkpointing":[83,109,177],"into":[84],"single,":[86],"unified":[87],"operation.":[88],"By":[89],"offloading":[90],"footprint":[92],"executing":[96],"these":[97],"fused":[98],"tasks":[99],"directly":[100],"within":[101,126],"internal":[102],"computational":[103],"units,":[104],"JBOC":[105,145,172,185],"achieves":[106,157],"highly":[107],"efficient":[108],"persistence":[111],"handled":[113],"during":[114],"update":[116],"pipeline.":[117],"offloaded":[119],"approach":[120],"allows":[121],"remain":[125],"deviceside":[127],"resources,":[128],"bypassing":[129],"host-side":[131],"resource":[132],"hierarchy":[133],"effectively":[135],"isolating":[136],"host":[138,198],"from":[139,150],"heavy":[140],"storage":[141],"traffic.":[142],"We":[143],"validate":[144],"commercial":[147],"CXL":[148],"platforms":[149],"Marvell":[151],"SK":[153],"Hynix.":[154],"Our":[155],"evaluation":[156],"up":[158],"<inline-formula":[161,178],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[162,179],"xmlns:xlink=\"http://www.w3.org/1999/xlink\"><tex-math":[163,180],"notation=\"LaTeX\">$4.40\\times$</tex-math></inline-formula>":[164],"speedup":[165],"under":[166],"per-iteration":[167],"Compared":[169],"ZeRO-Offload,":[171],"delivers":[173],"comparable":[174],"throughput":[175],"while":[176],"notation=\"LaTeX\">$10\\times$</tex-math></inline-formula>":[181],"more":[182],"frequently.":[183],"Furthermore,":[184],"outperforms":[186],"both":[187],"ZeROOffload":[188],"ZeRO-Infinity":[190],"in":[191],"memory-constrained":[192],"scenarios,":[193],"demonstrating":[194],"its":[195],"robustness":[196],"when":[197],"resources":[199],"are":[200],"limited.":[201]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-03-21T00:00:00"}
