{"id":"https://openalex.org/W4412623606","doi":"https://doi.org/10.1109/lca.2025.3592563","title":"SSD Offloading for LLM Mixture-of-Experts Weights Considered Harmful in Energy Efficiency","display_name":"SSD Offloading for LLM Mixture-of-Experts Weights Considered Harmful in Energy Efficiency","publication_year":2025,"publication_date":"2025-07-01","ids":{"openalex":"https://openalex.org/W4412623606","doi":"https://doi.org/10.1109/lca.2025.3592563"},"language":"en","primary_location":{"id":"doi:10.1109/lca.2025.3592563","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2025.3592563","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":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2508.06978","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5002219153","display_name":"Kwanhee Kyung","orcid":"https://orcid.org/0000-0003-4243-2111"},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":true,"raw_author_name":"Kwanhee Kyung","raw_affiliation_strings":["Seoul National University, Seoul, South Korea"],"affiliations":[{"raw_affiliation_string":"Seoul National University, Seoul, South Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5081681722","display_name":"Sungmin Yun","orcid":"https://orcid.org/0000-0002-5682-2535"},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sungmin Yun","raw_affiliation_strings":["Seoul National University, Seoul, South Korea"],"affiliations":[{"raw_affiliation_string":"Seoul National University, Seoul, South Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5078262826","display_name":"Jung Ho Ahn","orcid":"https://orcid.org/0000-0003-1733-1394"},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jung Ho Ahn","raw_affiliation_strings":["Seoul National University, Seoul, South Korea"],"affiliations":[{"raw_affiliation_string":"Seoul National University, Seoul, South Korea","institution_ids":["https://openalex.org/I139264467"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5002219153"],"corresponding_institution_ids":["https://openalex.org/I139264467"],"apc_list":null,"apc_paid":null,"fwci":1.5458,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.86172986,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":"24","issue":"2","first_page":"265","last_page":"268"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11181","display_name":"Advanced Data Storage Technologies","score":0.9994000196456909,"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":0.9994000196456909,"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/T11478","display_name":"Caching and Content Delivery","score":0.9940999746322632,"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/T10715","display_name":"Distributed and Parallel Computing Systems","score":0.9656999707221985,"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":0.7545431852340698},{"id":"https://openalex.org/keywords/energy-consumption","display_name":"Energy consumption","score":0.49428462982177734},{"id":"https://openalex.org/keywords/efficient-energy-use","display_name":"Efficient energy use","score":0.48796314001083374},{"id":"https://openalex.org/keywords/energy","display_name":"Energy (signal processing)","score":0.4119487702846527},{"id":"https://openalex.org/keywords/process-engineering","display_name":"Process engineering","score":0.38544154167175293},{"id":"https://openalex.org/keywords/risk-analysis","display_name":"Risk analysis (engineering)","score":0.3627663850784302},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.13136345148086548},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.11570608615875244},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.0965917706489563}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7545431852340698},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.49428462982177734},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.48796314001083374},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.4119487702846527},{"id":"https://openalex.org/C21880701","wikidata":"https://www.wikidata.org/wiki/Q2144042","display_name":"Process engineering","level":1,"score":0.38544154167175293},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.3627663850784302},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.13136345148086548},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.11570608615875244},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.0965917706489563},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.0},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/lca.2025.3592563","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2025.3592563","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"},{"id":"pmh:oai:arXiv.org:2508.06978","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2508.06978","pdf_url":"https://arxiv.org/pdf/2508.06978","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2508.06978","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2508.06978","pdf_url":"https://arxiv.org/pdf/2508.06978","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[{"score":0.8600000143051147,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":7,"referenced_works":["https://openalex.org/W2761132374","https://openalex.org/W3144954235","https://openalex.org/W4385192563","https://openalex.org/W4401211627","https://openalex.org/W4402671659","https://openalex.org/W4404955001","https://openalex.org/W6772383348"],"related_works":["https://openalex.org/W179829755","https://openalex.org/W4214628662","https://openalex.org/W4297745244","https://openalex.org/W2037518538","https://openalex.org/W2156822401","https://openalex.org/W3092380670","https://openalex.org/W3125847301","https://openalex.org/W4313390786","https://openalex.org/W2089393798","https://openalex.org/W2356819012"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"applying":[4],"Mixture-of-Experts":[5],"(MoE)":[6],"scale":[7],"to":[8,21,29,62,92,103,110],"trillions":[9],"of":[10,19,48,57,69,146,166],"parameters":[11],"but":[12],"require":[13],"vast":[14],"memory,":[15],"motivating":[16],"a":[17],"line":[18],"research":[20],"offload":[22],"expert":[23,60],"weights":[24,61,91],"from":[25],"fast-but-small":[26],"DRAM":[27],"(HBM)":[28],"denser":[30],"Flash":[31,157],"SSDs.":[32],"While":[33],"SSDs":[34,63,94,152],"provide":[35],"cost-effective":[36],"capacity,":[37],"their":[38],"read":[39,158],"energy":[40,55,98,118,133,159],"per":[41],"bit":[42],"is":[43],"substantially":[44],"higher":[45],"than":[46],"that":[47,88,142],"DRAM.":[49],"This":[50],"paper":[51],"quantitatively":[52],"analyzes":[53],"the":[54,65,111,115,143],"implications":[56],"offloading":[58,89],"MoE":[59,90,147],"during":[64],"critical":[66],"decode":[67],"stage":[68],"LLM":[70],"inference.":[71],"Our":[72],"analysis,":[73],"comparing":[74],"SSD,":[75],"CPU":[76],"memory":[77],"(DDR),":[78],"and":[79],"HBM":[80,112],"storage":[81],"scenarios":[82],"for":[83],"models":[84,148],"like":[85,122],"DeepSeek-R1,":[86],"reveals":[87],"current":[93],"drastically":[95],"increases":[96],"per-token-generation":[97],"consumption":[99],"(e.g.,":[100],"by":[101,163],"up":[102],"<inline-formula":[104],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[105,155],"xmlns:xlink=\"http://www.w3.org/1999/xlink\"><tex-math":[106],"notation=\"LaTeX\">$\\sim":[107],"12\\times$</tex-math></inline-formula>":[108],"compared":[109],"baseline),":[113],"dominating":[114],"total":[116],"inference":[117],"budget.":[119],"Although":[120],"techniques":[121],"prefetching":[123],"effectively":[124],"hide":[125],"access":[126],"latency,":[127],"they":[128],"cannot":[129],"mitigate":[130],"this":[131],"fundamental":[132],"penalty.":[134],"We":[135],"further":[136],"explore":[137],"future":[138],"technological":[139],"scaling,":[140],"finding":[141],"inherent":[144],"sparsity":[145],"could":[149],"potentially":[150],"make":[151],"energy-viable":[153],"<italic":[154],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">if</i>":[156],"improves":[160],"significantly,":[161],"roughly":[162],"an":[164],"order":[165],"magnitude.":[167]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}
