{"id":"https://openalex.org/W4413213050","doi":"https://doi.org/10.1109/lca.2025.3597323","title":"AiDE: Attention-FFN Disaggregated Execution for Cost-Effective LLM Decoding on CXL-PNM","display_name":"AiDE: Attention-FFN Disaggregated Execution for Cost-Effective LLM Decoding on CXL-PNM","publication_year":2025,"publication_date":"2025-07-01","ids":{"openalex":"https://openalex.org/W4413213050","doi":"https://doi.org/10.1109/lca.2025.3597323"},"language":"en","primary_location":{"id":"doi:10.1109/lca.2025.3597323","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2025.3597323","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/A5100719028","display_name":"Kyung-Soo Kim","orcid":"https://orcid.org/0000-0002-8927-1530"},"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":"KyungSoo Kim","raw_affiliation_strings":["Seoul National University, Seoul, South Korea","Seoul National University, Gwanak-gu, Seoul, Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0002-8927-1530","affiliations":[{"raw_affiliation_string":"Seoul National University, Seoul, South Korea","institution_ids":["https://openalex.org/I139264467"]},{"raw_affiliation_string":"Seoul National University, Gwanak-gu, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Omin Kwon","orcid":"https://orcid.org/0009-0007-7468-0810"},"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":"Omin Kwon","raw_affiliation_strings":["Seoul National University, Seoul, South Korea","Seoul National University, Gwanak-gu, Seoul, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0007-7468-0810","affiliations":[{"raw_affiliation_string":"Seoul National University, Seoul, South Korea","institution_ids":["https://openalex.org/I139264467"]},{"raw_affiliation_string":"Seoul National University, Gwanak-gu, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049560946","display_name":"Yeonhong Park","orcid":"https://orcid.org/0009-0008-1425-0971"},"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":"Yeonhong Park","raw_affiliation_strings":["Seoul National University, Seoul, South Korea","Seoul National University, Gwanak-gu, Seoul, Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Seoul National University, Seoul, South Korea","institution_ids":["https://openalex.org/I139264467"]},{"raw_affiliation_string":"Seoul National University, Gwanak-gu, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100415731","display_name":"Jae W. Lee","orcid":"https://orcid.org/0000-0002-4266-4919"},"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":"Jae W. Lee","raw_affiliation_strings":["Seoul National University, Seoul, South Korea","Seoul National University, Gwanak-gu, Seoul, Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0002-4266-4919","affiliations":[{"raw_affiliation_string":"Seoul National University, Seoul, South Korea","institution_ids":["https://openalex.org/I139264467"]},{"raw_affiliation_string":"Seoul National University, Gwanak-gu, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I139264467"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5100719028"],"corresponding_institution_ids":["https://openalex.org/I139264467"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.2215494,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"24","issue":"2","first_page":"285","last_page":"288"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11181","display_name":"Advanced Data Storage Technologies","score":0.866100013256073,"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":0.866100013256073,"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/T11032","display_name":"VLSI and Analog Circuit Testing","score":0.8619999885559082,"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/T11301","display_name":"Advanced Surface Polishing Techniques","score":0.8373000025749207,"subfield":{"id":"https://openalex.org/subfields/2204","display_name":"Biomedical 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/decoding-methods","display_name":"Decoding methods","score":0.7812991738319397},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7564650774002075},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.4191686511039734},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.19586700201034546}],"concepts":[{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.7812991738319397},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7564650774002075},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4191686511039734},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.19586700201034546}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/lca.2025.3597323","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2025.3597323","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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":7,"referenced_works":["https://openalex.org/W4385245566","https://openalex.org/W4389518760","https://openalex.org/W4392427708","https://openalex.org/W4393407316","https://openalex.org/W4394998968","https://openalex.org/W4401211704","https://openalex.org/W4407425079"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Disaggregating":[0],"the":[1,14,23,41,45,123],"prefill":[2],"and":[3,58,101,145],"decode":[4,42,150],"phases":[5],"has":[6],"recently":[7],"emerged":[8],"as":[9],"a":[10,37,73,113,146,154],"promising":[11],"strategy":[12],"in":[13,149],"large":[15],"language":[16],"model":[17],"(LLM)":[18],"serving":[19],"systems,":[20],"driven":[21],"by":[22,31,55],"distinct":[24],"resource":[25],"demands":[26],"of":[27,125,163],"each":[28],"phase.":[29],"Inspired":[30],"this":[32,68],"coarse-grained":[33],"disaggregation,":[34],"we":[35,70,111],"identify":[36],"similar":[38],"opportunity":[39],"within":[40],"phase":[43],"itself:":[44],"feedforward":[46],"network":[47],"(FFN)":[48],"is":[49,53],"compute-intensive,":[50],"whereas":[51],"attention":[52,85],"constrained":[54],"memory":[56,99],"bandwidth":[57,100],"capacity":[59],"due":[60],"to":[61,87,121,133,153],"its":[62],"key-value":[63],"(KV)":[64],"cache.":[65],"To":[66],"exploit":[67],"heterogeneity,":[69],"introduce":[71],"AiDE,":[72],"heterogeneous":[74,126],"decoding":[75],"cluster":[76],"that":[77],"executes":[78],"FFN":[79],"operations":[80],"on":[81],"GPUs":[82],"while":[83],"offloading":[84],"computations":[86],"Compute":[88],"Express":[89],"Link-based":[90],"Processing":[91],"Near":[92],"Memory":[93],"(CXL-PNM)":[94],"devices.":[95],"CXL-PNM":[96],"provides":[97],"scalable":[98],"capacity,":[102],"making":[103],"it":[104],"well-suited":[105],"for":[106,166],"attention-heavy":[107],"workloads.":[108],"In":[109],"addition,":[110],"propose":[112],"batch-level":[114],"pipelining":[115],"approach":[116],"enhanced":[117],"with":[118],"request":[119],"scheduling":[120],"optimize":[122],"utilization":[124],"resources.":[127],"Our":[128],"AiDE":[129],"architecture":[130],"achieves":[131],"up":[132],"3.87\u00d7":[134],"higher":[135],"throughput,":[136],"2.72\u00d7":[137],"lower":[138],"p90":[139],"time":[140],"per":[141],"output":[142],"token":[143],"(TPOT),":[144],"2.31\u00d7":[147],"reduction":[148],"latency":[151],"compared":[152],"GPU-only":[155],"baseline,":[156],"at":[157],"comparable":[158],"cost,":[159],"demonstrating":[160],"significant":[161],"potential":[162],"fine-grained":[164],"disaggregation":[165],"cost-effective":[167],"LLM":[168],"inference.":[169]},"counts_by_year":[],"updated_date":"2025-12-19T19:40:27.379048","created_date":"2025-10-10T00:00:00"}