{"id":"https://openalex.org/W7225412337","doi":"https://doi.org/10.5281/zenodo.23997595","title":"Predictive GPU Memory Fragmentation Management for Generative AI: Forecasting Future Allocation Fragmentation and Anticipating Computational Resource Waste","display_name":"Predictive GPU Memory Fragmentation Management for Generative AI: Forecasting Future Allocation Fragmentation and Anticipating Computational Resource Waste","publication_year":2026,"publication_date":"2026-11-06","ids":{"openalex":"https://openalex.org/W7225412337","doi":"https://doi.org/10.5281/zenodo.23997595"},"language":"en","primary_location":{"id":"doi:10.5281/zenodo.23997595","doi":"https://doi.org/10.5281/zenodo.23997595","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23997595","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.5281/zenodo.23997595","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100638045","display_name":"Jincheng Zhang","orcid":"https://orcid.org/0000-0001-7332-6704","observed_orcids":["https://orcid.org/0000-0001-7332-6704","https://orcid.org/0009-0005-1860-0009"]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Zhang, Jincheng","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0009-0005-1860-0009","affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5100638045"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9911999702453613,"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"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9911999702453613,"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/T10101","display_name":"Cloud Computing and Resource Management","score":0.0024999999441206455,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10317","display_name":"Advanced Database Systems and Queries","score":0.0005000000237487257,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/large-language-models","display_name":"large language models","score":0.9169999957084656},{"id":"https://openalex.org/keywords/memory-reorganization","display_name":"memory reorganization","score":0.5860000252723694},{"id":"https://openalex.org/keywords/dynamic-batching","display_name":"dynamic batching","score":0.8069999814033508},{"id":"https://openalex.org/keywords/memory-pooling","display_name":"memory pooling","score":0.703000009059906}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8388000130653381},{"id":"https://openalex.org/C3720319","wikidata":"https://www.wikidata.org/wiki/Q5015937","display_name":"Cache-only memory architecture","level":5,"score":0.5809000134468079},{"id":"https://openalex.org/C176649486","wikidata":"https://www.wikidata.org/wiki/Q2308807","display_name":"Memory management","level":3,"score":0.5742999911308289},{"id":"https://openalex.org/C57863822","wikidata":"https://www.wikidata.org/wiki/Q905488","display_name":"Flat memory model","level":4,"score":0.5508999824523926},{"id":"https://openalex.org/C191015642","wikidata":"https://www.wikidata.org/wiki/Q1132459","display_name":"Fragmentation (computing)","level":2,"score":0.48579999804496765},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.47510001063346863},{"id":"https://openalex.org/C51290061","wikidata":"https://www.wikidata.org/wiki/Q1936765","display_name":"Uniform memory access","level":4,"score":0.47380000352859497},{"id":"https://openalex.org/C63511323","wikidata":"https://www.wikidata.org/wiki/Q908936","display_name":"Interleaved memory","level":4,"score":0.45590001344680786},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.45260000228881836},{"id":"https://openalex.org/C133875982","wikidata":"https://www.wikidata.org/wiki/Q764810","display_name":"Shared memory","level":2,"score":0.4456000030040741},{"id":"https://openalex.org/C29202148","wikidata":"https://www.wikidata.org/wiki/Q287260","display_name":"Resource allocation","level":2,"score":0.4106000065803528},{"id":"https://openalex.org/C74426580","wikidata":"https://www.wikidata.org/wiki/Q719484","display_name":"Memory map","level":3,"score":0.38499999046325684},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.35690000653266907},{"id":"https://openalex.org/C12186640","wikidata":"https://www.wikidata.org/wiki/Q6815743","display_name":"Memory model","level":3,"score":0.3479999899864197},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.3449999988079071},{"id":"https://openalex.org/C118702147","wikidata":"https://www.wikidata.org/wiki/Q189396","display_name":"Dynamic random-access memory","level":3,"score":0.3391000032424927},{"id":"https://openalex.org/C152890283","wikidata":"https://www.wikidata.org/wiki/Q4129922","display_name":"Computing with Memory","level":5,"score":0.3292999863624573},{"id":"https://openalex.org/C82687282","wikidata":"https://www.wikidata.org/wiki/Q66221","display_name":"Auxiliary memory","level":2,"score":0.3287999927997589},{"id":"https://openalex.org/C133371097","wikidata":"https://www.wikidata.org/wiki/Q868014","display_name":"Non-uniform memory access","level":5,"score":0.32269999384880066},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.32019999623298645},{"id":"https://openalex.org/C34339311","wikidata":"https://www.wikidata.org/wiki/Q1050390","display_name":"C dynamic memory allocation","level":4,"score":0.3158999979496002},{"id":"https://openalex.org/C91481028","wikidata":"https://www.wikidata.org/wiki/Q1054686","display_name":"Distributed memory","level":3,"score":0.3107999861240387},{"id":"https://openalex.org/C171675096","wikidata":"https://www.wikidata.org/wiki/Q1143380","display_name":"Extended memory","level":4,"score":0.29100000858306885},{"id":"https://openalex.org/C49154492","wikidata":"https://www.wikidata.org/wiki/Q5300","display_name":"Central processing unit","level":2,"score":0.28049999475479126},{"id":"https://openalex.org/C30390489","wikidata":"https://www.wikidata.org/wiki/Q4680748","display_name":"Adaptive memory","level":3,"score":0.27970001101493835},{"id":"https://openalex.org/C2779602883","wikidata":"https://www.wikidata.org/wiki/Q15544750","display_name":"Memory architecture","level":2,"score":0.2721000015735626},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.27070000767707825},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.26899999380111694},{"id":"https://openalex.org/C2781357197","wikidata":"https://www.wikidata.org/wiki/Q5757597","display_name":"High memory","level":2,"score":0.266400009393692},{"id":"https://openalex.org/C162262903","wikidata":"https://www.wikidata.org/wiki/Q343527","display_name":"Allocator","level":2,"score":0.2651999890804291}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.5281/zenodo.23997595","doi":"https://doi.org/10.5281/zenodo.23997595","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23997595","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},{"id":"doi:10.5281/zenodo.23997596","doi":"https://doi.org/10.5281/zenodo.23997596","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23997596","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.5281/zenodo.23997595","doi":"https://doi.org/10.5281/zenodo.23997595","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23997595","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7225392849","https://openalex.org/W7225424338","https://openalex.org/W7225432460","https://openalex.org/W7225409378","https://openalex.org/W4416763335","https://openalex.org/W7168016844","https://openalex.org/W7225405737","https://openalex.org/W7225408802","https://openalex.org/W7225390238","https://openalex.org/W7225400169"],"abstract_inverted_index":{"The":[0,230,298],"increasing":[1],"deployment":[2],"of":[3,91,101,163,170,179,227,307,338],"large":[4],"language":[5],"models":[6],"(LLMs)":[7],"and":[8,45,62,81,131,165,176,200,221,249,294,335],"other":[9],"generative":[10,24,368],"artificial":[11],"intelligence":[12],"(AI)":[13],"applications":[14],"has":[15],"introduced":[16],"increasingly":[17],"complex":[18],"GPU":[19,51,93,116,128,145,308,364],"memory":[20,29,52,58,70,74,94,102,129,146,153,172,198,216,220,245,261,287,290,292,309,330,348,365],"allocation":[21,80,140,181,194,251,273,296],"patterns.":[22],"Modern":[23],"AI":[25,369],"workloads":[26],"involve":[27],"heterogeneous":[28,250],"requirements":[30,350],"associated":[31,132],"with":[32,285],"model":[33,256],"parameters,":[34],"intermediate":[35],"activations,":[36],"attention":[37],"operations,":[38],"key-value":[39],"(KV)":[40],"caches,":[41,242],"temporary":[42],"tensors,":[43],"batching,":[44,244],"concurrent":[46,237],"inference":[47,238],"requests.":[48],"Although":[49],"existing":[50],"management":[53,66,284],"techniques":[54],"employ":[55],"caching":[56],"allocators,":[57],"pooling,":[59,293],"dynamic":[60,243],"allocation,":[61,288],"specialized":[63],"KV":[64,241],"cache":[65],"mechanisms":[67],"to":[68,108,259,280,351],"improve":[69],"utilization,":[71],"the":[72,88,99,159,168,177,223,234,336],"resulting":[73],"layout":[75],"may":[76],"become":[77],"fragmented":[78],"as":[79],"deallocation":[82],"operations":[83],"accumulate":[84],"over":[85],"time.":[86],"Consequently,":[87],"total":[89,214],"amount":[90,100],"available":[92,166,265],"does":[95,301],"not":[96,302],"necessarily":[97],"represent":[98],"that":[103,124,212,262,304,315],"can":[104,135,311],"be":[105,136,268,312],"effectively":[106,218,269],"allocated":[107,164],"future":[109,127,152,160,180,191,228,353],"computational":[110,225],"tasks.":[111],"This":[112],"paper":[113],"introduces":[114],"Predictive":[115],"Memory":[117],"Fragmentation":[118],"Management":[119],"(PGMFM),":[120],"a":[121,208,359],"conceptual":[122,360],"framework":[123,231],"investigates":[125],"whether":[126],"fragmentation":[130,189,283,310],"resource":[133,202],"waste":[134],"anticipated":[137,272],"before":[138],"subsequent":[139],"requests":[141],"materialize.":[142],"Unlike":[143],"conventional":[144,286],"demand":[147],"forecasting,":[148,196],"which":[149],"primarily":[150],"estimates":[151,222],"consumption,":[154],"PGMFM":[155],"focuses":[156],"on":[157,325],"predicting":[158],"spatial":[161],"distribution":[162],"memory,":[167],"formation":[169],"unusable":[171],"regions,":[173],"allocation-size":[174],"incompatibility,":[175],"probability":[178],"failures.":[182],"Its":[183,321],"architecture":[184],"integrates":[185],"allocation-history":[186],"monitoring,":[187],"workload-aware":[188,295],"prediction,":[190],"memory-layout":[192,354],"estimation,":[193],"feasibility":[195],"predictive":[197,282,316,344],"reorganization,":[199],"adaptive":[201],"management.":[203],"A":[204,253],"central":[205],"component":[206],"is":[207,257,278,318],"fragmentation-aware":[209],"optimization":[210],"formulation":[211],"distinguishes":[213],"free":[215],"from":[217,346],"allocatable":[219],"expected":[224],"cost":[226],"fragmentation.":[229],"further":[232],"considers":[233],"interaction":[235],"between":[236],"requests,":[239],"variable-length":[240],"retention,":[246],"tensor":[247],"lifetimes,":[248],"sizes.":[252],"resource-waste":[254],"prediction":[255],"developed":[258],"characterize":[260],"remains":[263],"physically":[264],"but":[266],"cannot":[267],"utilized":[270],"under":[271],"requirements.":[274],"An":[275],"experimental":[276],"methodology":[277],"proposed":[279,299],"compare":[281],"reactive":[289],"defragmentation,":[291],"strategies.":[297],"research":[300,357],"assume":[303],"all":[305],"forms":[306],"eliminated":[313],"or":[314],"reorganization":[317],"universally":[319],"beneficial.":[320],"practical":[322],"effectiveness":[323],"depends":[324],"workload":[326],"predictability,":[327],"allocator":[328],"behavior,":[329],"relocation":[331],"constraints,":[332],"execution":[333],"overhead,":[334],"availability":[337],"compatible":[339],"memory-management":[340],"mechanisms.":[341],"By":[342],"extending":[343],"computing":[345],"forecasting":[347],"capacity":[349],"anticipating":[352],"inefficiencies,":[355],"this":[356],"establishes":[358],"foundation":[361],"for":[362],"proactive":[363],"organization":[366],"in":[367],"infrastructure.":[370]},"counts_by_year":[],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2026-11-07T00:00:00"}