Article(id=1327011159850463496, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327011123016085523, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202440628, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1721404800000, receivedDateStr=2024-07-20, revisedDate=1737302400000, revisedDateStr=2025-01-20, acceptedDate=null, acceptedDateStr=null, onlineDate=1794313715183, onlineDateStr=2026-11-10, pubDate=1740758400000, pubDateStr=2025-03-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794313715183, onlineIssueDateStr=2026-11-10, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794313715183, creator=13701087609, updateTime=1794313715183, updator=13701087609, issue=Issue{id=1327011123016085523, tenantId=1146029695717560320, journalId=1324759914146529478, year='2025', volume='62', issue='3', pageStart='543', pageEnd='819', issueExtLink='null', onlineDate='null', pubDate='1740758400000', pubDateStr='2025-03-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794313706401, creator='13701087609', updateTime=1794376726899, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1327275450650440506, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327011123016085523, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1327275450650440507, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1327011123016085523, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=545, endPage=562, ext={EN=ArticleExt(id=1327011160345391369, articleId=1327011159850463496, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Survey of Storage Optimization Techniques in Large Language Model Inference, columnId=1327011123649425428, journalTitle=Journal of Computer Research and Development, columnName=Special Issue on Storage Management and Data Analysis in the Era of Large Language Models, runingTitle=null, highlight=null, articleAbstract=
In recent years, LLM (large language model) has exhibited remarkable performance, profoundly transforming various aspects of human life. As these models grow in size and user demand for long-context inference increases, LLM inference systems face significant storage challenges. These challenges stem primarily from the vast number of model parameters and the key value cache required for efficient inference, both of which strain GPU memory resources. Additionally, inefficiencies in storage usage in distributed systems often result in over-provisioning and fault tolerance issues, further complicating resource management. Researchers explore memory optimization, heterogeneous storage, and distributed storage, synthesizing various research efforts to address GPU memory constraints and enhance resource utilization. Memory-optimized LLM inference systems improve GPU memory efficiency and reduce memory footprint through techniques like efficient key value cache management, compression, and attention operator optimization. Heterogeneous storage based LLM inference systems expand storage capacity by integrating various storage resources, thereby minimizing I/O overhead via tensor placement strategies, asynchronous data transfer, and intelligent memory allocation and prefetching methods. Distributed LLM systems enhance the utilization of multi-machine resources, boosting execution efficiency and fault tolerance in LLM inference tasks through batching, multi-level scheduling, and redundant replication. Finally, we review existing research and outline future research directions to further optimize storage solutions for LLM inference systems.
, authors=Xuran Ge
1, Yang Ou
1, *, Bo Wang
2, Yu Zhao
3, Lizhou Wu
1, Zicong Wang
1, Zhiguang Chen
2, Nong Xiao
1, authorsList=Xuran Ge, Yang Ou, Bo Wang, Yu Zhao, Lizhou Wu, Zicong Wang, Zhiguang Chen, Nong Xiao, authorCompany=null, correspAuthors=Yang Ou, authorNote=
born in 1998. PhD candidate. Student member of CCF. Her main research interests include key-value storage system and large language model inference system
born in 1987. PhD, associate professor. His main research interests include computer architecture and high performance interconnection technology
born in 1997. PhD candidate. Student member of CCF. His main research interests include distributed shared memory and parallel programming frameworks
born in 1997. PhD candidate. His main research interest includes efficient inference and pre-training of large language model
born in 1991. PhD, assistant professor. His main research interests include high performance computing, spintronic design and test, and emerging computing paradigms based on non-volatile memory devices
born in 1991. PhD, assistant professor. His main research interests include computer architecture and high performance interconnection technology
born in 1984. PhD, professor,PhD supervisor. Member of CCF. His main research interests include big data storage, and high performance computing and supercomputers
born in 1969. PhD, professor,PhD supervisor. His main research interests include high performance computing, storage system, and system architecture
, correspAuthorsNote=null, copyrightStatement=Copyright © 2025 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1327011168780136769, articleId=1327011159850463496, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=大语言模型推理中的存储优化技术综述, columnId=1327011123825586198, journalTitle=计算机研究与发展, columnName=大模型时代的存储管理与数据分析专题, runingTitle=null, highlight=null, articleAbstract=
近年来,大语言模型在多个领域展现出卓越的性能,显著改变了人类的生活方式. 然而,随着模型规模的不断增长和用户对长上下文推理需求的增加,大语言模型推理系统在存储方面面临诸多问题. 首先,庞大的模型参数和键值缓存导致GPU显存资源不足;其次,分布式大语言模型推理系统难以充分利用GPU集群的存储资源,存在资源过度配置和存储容错的问题. 因此,从显存优化、异构存储和分布式存储3方面入手,归纳总结了现有研究在解决GPU显存容量不足和资源利用率低下方面的努力. 基于显存优化的大语言模型推理系统通过高效的键值缓存管理、压缩以及注意力算子优化,提高了GPU显存利用率,降低了显存占用. 基于异构存储的大语言模型推理系统利用多种类别的存储资源扩展存储容量,通过张量放置策略、异步数据传输以及智能显存分配与预取技术,降低了异构存储带来的I/O开销. 基于分布式存储的大语言模型推理系统通过批处理、多级调度、冗余复制等策略,优化了多机存储和计算资源的利用,提高了大语言模型推理任务的执行效率和容错能力. 最后,总结了现有研究,并对未来的研究方向进行了展望.
, authors=葛旭冉
1, 欧洋
1, *, 王博
2, 赵宇
3, 吴利舟
1, 王子聪
1, 陈志广
2, 肖侬
1, authorsList=葛旭冉, 欧洋, 王博, 赵宇, 吴利舟, 王子聪, 陈志广, 肖侬, authorCompany=null, correspAuthors=欧洋, authorNote=
 |
1998年生. 博士研究生. CCF学生会员. 主要研究方向为键值存储系统、大语言模型推理系统 |
 |
1987年生. 博士,副研究员. 主要研究方向为计算机系统结构、高性能互联技术 |
 |
1997年生. 博士研究生. CCF学生会员. 主要研究方向为分布式共享内存、并行编程框架 |
 |
1997年生. 博士研究生. 主要研究方向为大语言模型高效推理及预训练 |
 |
1991年生. 博士,助理研究员. 主要研究方向为高性能计算、自旋电子设计与测试、基于新型非易失存储器的计算范式 |
 |
1991年生. 博士,助理研究员. 主要研究方向为计算机系统结构、高性能互联技术 |
 |
1984年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为大数据存储、高性能计算与超级计算机 |
 |
1969年生. 博士,教授,博士生导师. 主要研究方向为高性能计算、存储系统、系统结构 |
, correspAuthorsNote=
, copyrightStatement=版权所有 © 2025 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=JYFc+XqmkA0IwFsQkm7tdQ==, magXml=csOOjj15arVkXduSx4ZOWA==, pdfUrl=null, pdf=HYrIIjHtOhTd3SB+8/+nzw==, pdfFileSize=2424631, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=hUlcZ8WpR7t7IbklUh6OTA==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=t5QiuHTrpY2IPdwc02ZOew==, mapNumber=null, fund=null)}, authors=[Author(id=1327011169627386188, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=gexuran@nudt.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327011169723855182, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011169627386188, language=EN, stringName=Xuran Ge, firstName=Xuran, middleName=null, lastName=Ge, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, address=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"
born in 1998. PhD candidate. Student member of CCF. Her main research interests include key-value storage system and large language model inference system
"}, bioImg=null, bioContent=
born in 1998. PhD candidate. Student member of CCF. Her main research interests include key-value storage system and large language model inference system
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327011169795158351, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011169627386188, language=CN, stringName=葛旭冉, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, address=
1国防科技大学计算机学院 长沙 410073, bio={"img":"n/n2ZPW5qR/6N/zbdNonTQ==","content":"
1998年生. 博士研究生. CCF学生会员. 主要研究方向为键值存储系统、大语言模型推理系统
"}, bioImg=n/n2ZPW5qR/6N/zbdNonTQ==, bioContent=
1998年生. 博士研究生. CCF学生会员. 主要研究方向为键值存储系统、大语言模型推理系统
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327011169279258946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=1, ext=[AuthorCompanyExt(id=1327011169304424771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1327011169337979204, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1国防科技大学计算机学院 长沙 410073)])]), Author(id=1327011169874850129, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=ouyang06@nudt.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1327011169988096339, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011169874850129, language=EN, stringName=Yang Ou, firstName=Yang, middleName=null, lastName=Ou, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, *, address=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"
born in 1987. PhD, associate professor. His main research interests include computer architecture and high performance interconnection technology
"}, bioImg=null, bioContent=
born in 1987. PhD, associate professor. His main research interests include computer architecture and high performance interconnection technology
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327011170051010900, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011169874850129, language=CN, stringName=欧洋, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, *, address=
1国防科技大学计算机学院 长沙 410073, bio={"img":"qI5rp/f9Jfwbkc0P0jL5ig==","content":"
1987年生. 博士,副研究员. 主要研究方向为计算机系统结构、高性能互联技术
"}, bioImg=qI5rp/f9Jfwbkc0P0jL5ig==, bioContent=
1987年生. 博士,副研究员. 主要研究方向为计算机系统结构、高性能互联技术
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327011169279258946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=1, ext=[AuthorCompanyExt(id=1327011169304424771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1327011169337979204, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1国防科技大学计算机学院 长沙 410073)])]), Author(id=1327011171732926806, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327011171812618584, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011171732926806, language=EN, stringName=Bo Wang, firstName=Bo, middleName=null, lastName=Wang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
2, address=
2School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006, bio={"content":"
born in 1997. PhD candidate. Student member of CCF. His main research interests include distributed shared memory and parallel programming frameworks
"}, bioImg=null, bioContent=
born in 1997. PhD candidate. Student member of CCF. His main research interests include distributed shared memory and parallel programming frameworks
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327011171892310361, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011171732926806, language=CN, stringName=王博, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
2, address=
2中山大学计算机学院 广州 510006, bio={"img":"d4iBzL1FdWcAoSl7nQpdqw==","content":"
1997年生. 博士研究生. CCF学生会员. 主要研究方向为分布式共享内存、并行编程框架
"}, bioImg=d4iBzL1FdWcAoSl7nQpdqw==, bioContent=
1997年生. 博士研究生. CCF学生会员. 主要研究方向为分布式共享内存、并行编程框架
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327011169434448197, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=2, ext=[AuthorCompanyExt(id=1327011169442836806, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169434448197, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
2School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006), AuthorCompanyExt(id=1327011169451225415, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169434448197, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
2中山大学计算机学院 广州 510006)])]), Author(id=1327011172001362267, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327011172081054045, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172001362267, language=EN, stringName=Yu Zhao, firstName=Yu, middleName=null, lastName=Zhao, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
3, address=
3School of Informatics, University of Edinburgh, Edinburgh EH89AB, bio={"content":"
born in 1997. PhD candidate. His main research interest includes efficient inference and pre-training of large language model
"}, bioImg=null, bioContent=
born in 1997. PhD candidate. His main research interest includes efficient inference and pre-training of large language model
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327011172198494558, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172001362267, language=CN, stringName=赵宇, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
3, address=
3爱丁堡大学信息学院 爱丁堡 EH89AB, bio={"img":"ljOf5OUkmzgcgd3Cgm92mA==","content":"
1997年生. 博士研究生. 主要研究方向为大语言模型高效推理及预训练
"}, bioImg=ljOf5OUkmzgcgd3Cgm92mA==, bioContent=
1997年生. 博士研究生. 主要研究方向为大语言模型高效推理及预训练
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327011169518334280, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=3, ext=[AuthorCompanyExt(id=1327011169530917193, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169518334280, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
3School of Informatics, University of Edinburgh, Edinburgh EH89AB), AuthorCompanyExt(id=1327011169539305802, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169518334280, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
3爱丁堡大学信息学院 爱丁堡 EH89AB)])]), Author(id=1327011172273992032, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327011172374655330, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172273992032, language=EN, stringName=Lizhou Wu, firstName=Lizhou, middleName=null, lastName=Wu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, address=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"
born in 1991. PhD, assistant professor. His main research interests include high performance computing, spintronic design and test, and emerging computing paradigms based on non-volatile memory devices
"}, bioImg=null, bioContent=
born in 1991. PhD, assistant professor. His main research interests include high performance computing, spintronic design and test, and emerging computing paradigms based on non-volatile memory devices
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327011172458541411, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172273992032, language=CN, stringName=吴利舟, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, address=
1国防科技大学计算机学院 长沙 410073, bio={"img":"srjye+qUH25W9j3jqqXlyA==","content":"
1991年生. 博士,助理研究员. 主要研究方向为高性能计算、自旋电子设计与测试、基于新型非易失存储器的计算范式
"}, bioImg=srjye+qUH25W9j3jqqXlyA==, bioContent=
1991年生. 博士,助理研究员. 主要研究方向为高性能计算、自旋电子设计与测试、基于新型非易失存储器的计算范式
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327011169279258946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=1, ext=[AuthorCompanyExt(id=1327011169304424771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1327011169337979204, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1国防科技大学计算机学院 长沙 410073)])]), Author(id=1327011172529844581, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, orderNo=5, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327011172617924967, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172529844581, language=EN, stringName=Zicong Wang, firstName=Zicong, middleName=null, lastName=Wang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, address=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"
born in 1991. PhD, assistant professor. His main research interests include computer architecture and high performance interconnection technology
"}, bioImg=null, bioContent=
born in 1991. PhD, assistant professor. His main research interests include computer architecture and high performance interconnection technology
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327011172693422440, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172529844581, language=CN, stringName=王子聪, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, address=
1国防科技大学计算机学院 长沙 410073, bio={"img":"2hb3M4yo/FE3+nl5afyRKg==","content":"
1991年生. 博士,助理研究员. 主要研究方向为计算机系统结构、高性能互联技术
"}, bioImg=2hb3M4yo/FE3+nl5afyRKg==, bioContent=
1991年生. 博士,助理研究员. 主要研究方向为计算机系统结构、高性能互联技术
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327011169279258946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=1, ext=[AuthorCompanyExt(id=1327011169304424771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1327011169337979204, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1国防科技大学计算机学院 长沙 410073)])]), Author(id=1327011172760531306, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, orderNo=6, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327011172844417388, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172760531306, language=EN, stringName=Zhiguang Chen, firstName=Zhiguang, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
2, address=
2School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006, bio={"content":"
born in 1984. PhD, professor,PhD supervisor. Member of CCF. His main research interests include big data storage, and high performance computing and supercomputers
"}, bioImg=null, bioContent=
born in 1984. PhD, professor,PhD supervisor. Member of CCF. His main research interests include big data storage, and high performance computing and supercomputers
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327011172911526253, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172760531306, language=CN, stringName=陈志广, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
2, address=
2中山大学计算机学院 广州 510006, bio={"img":"LZL84+54tlS7GX4pijDRtg==","content":"
1984年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为大数据存储、高性能计算与超级计算机
"}, bioImg=LZL84+54tlS7GX4pijDRtg==, bioContent=
1984年生. 博士,教授,博士生导师. CCF会员. 主要研究方向为大数据存储、高性能计算与超级计算机
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327011169434448197, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=2, ext=[AuthorCompanyExt(id=1327011169442836806, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169434448197, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
2School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006), AuthorCompanyExt(id=1327011169451225415, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169434448197, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
2中山大学计算机学院 广州 510006)])]), Author(id=1327011172974440815, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, orderNo=7, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1327011173045743985, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172974440815, language=EN, stringName=Nong Xiao, firstName=Nong, middleName=null, lastName=Xiao, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, address=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, bio={"content":"
born in 1969. PhD, professor,PhD supervisor. His main research interests include high performance computing, storage system, and system architecture
"}, bioImg=null, bioContent=
born in 1969. PhD, professor,PhD supervisor. His main research interests include high performance computing, storage system, and system architecture
, aboutCorrespAuthor=null), CN=AuthorExt(id=1327011173121241458, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, authorId=1327011172974440815, language=CN, stringName=肖侬, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=
1, address=
1国防科技大学计算机学院 长沙 410073, bio={"img":"qI4qKcNWml6zedksL23K5g==","content":"
1969年生. 博士,教授,博士生导师. 主要研究方向为高性能计算、存储系统、系统结构
"}, bioImg=qI4qKcNWml6zedksL23K5g==, bioContent=
1969年生. 博士,教授,博士生导师. 主要研究方向为高性能计算、存储系统、系统结构
, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1327011169279258946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=1, ext=[AuthorCompanyExt(id=1327011169304424771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1327011169337979204, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1国防科技大学计算机学院 长沙 410073)])])], keywords=[Keyword(id=1327011173217710451, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, orderNo=1, keyword=LLM inference system), Keyword(id=1327011173280625012, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, orderNo=2, keyword=memory management), Keyword(id=1327011173343539573, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, orderNo=3, keyword=heterogeneous storage), Keyword(id=1327011173406454134, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, orderNo=4, keyword=distributed storage), Keyword(id=1327011173473562999, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, orderNo=5, keyword=fault recovery), Keyword(id=1327011173532283256, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, orderNo=6, keyword=serverless LLM inference), Keyword(id=1327011173603586425, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, orderNo=1, keyword=大语言模型推理系统), Keyword(id=1327011173662306682, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, orderNo=2, keyword=显存管理), Keyword(id=1327011173725221243, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, orderNo=3, keyword=异构存储), Keyword(id=1327011173796524412, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, orderNo=4, keyword=分布式存储), Keyword(id=1327011173863633277, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, orderNo=5, keyword=故障恢复), Keyword(id=1327011173943325054, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, orderNo=6, keyword=无服务大语言模型推理)], refs=[Reference(id=1327011181019115947, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=OpenAI. GPT−4 technical report[J]. arXiv preprint, arXiv: 2303.08774, 2023, articleTitle=null, refAbstract=null), Reference(id=1327011181077836204, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Shuai, Roller S, Goyal N, et al. OPT: Open pre-trained transformer language models[J]. arXiv preprint, arXiv: 2205.01068, 2022, articleTitle=null, refAbstract=null), Reference(id=1327011181224636845, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Almazrouei E, Alobeidli H, Alshamsi A, et al. Falcon−40B: An open large language model with state-of-the-art performance[J]. arXiv preprint, arXiv: 2306.01116, 2023, articleTitle=null, refAbstract=null), Reference(id=1327011181308522926, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Touvron H, Lavril T, Izacard G, et al. LLaMA: Open and efficient foundation language models[J]. arXiv preprint, arXiv: 2302.13971, 2023, articleTitle=null, refAbstract=null), Reference(id=1327011181384020399, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=GitHub. Quantifying the productivity benefits of GitHub Copilot [EB/OL]. [2024-07-18]. https://github.com/features/copilot, articleTitle=null, refAbstract=null), Reference(id=1327011181446934960, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen M, Tworek J, Jun H, et al. Evaluating large language models trained on code[J]. arXiv preprint, arXiv: 2107.03374, 2023, articleTitle=null, refAbstract=null), Reference(id=1327011181514043825, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Du Jiangsu, Liu Ziming, Fang Jiarui, et al. EnergonAI: An inference system for 10−100 billion parameter transformer models[J]. arXiv preprint, arXiv: 2209.02341, 2022, articleTitle=null, refAbstract=null), Reference(id=1327011181597929906, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Villalobos P, Sevilla J, Besiroglu T, et al. Machine learning model sizes and the parameter gap[J]. arXiv preprint, arXiv: 2207.02852, 2022, articleTitle=null, refAbstract=null), Reference(id=1327011181677621683, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Brown T, Mann B, Ryder N, et al. Language models are few-shot learners [C]//Proc of the 34th Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2020: 1877−1901, articleTitle=null, refAbstract=null), Reference(id=1327011181761507764, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=NVIDIA. NVIDIA doubles down: Announces A100 80GB GPU, supercharging world’s most powerful GPU for AI supercomputing [EB/OL]. 2023[2024-07-18]. https://nvidianews.nvidia.com/news/nvidia-doubles-down-announces-a100-80gb-gpu-supercharging-worlds-most-powerful-gpu-for-ai-supercomputing, articleTitle=null, refAbstract=null), Reference(id=1327011181866365365, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Devoto A, Zhao Yu, Scardapane S, et al. A simple and effective L2 norm-based strategy for KV cache compression[J]. arXiv preprint, arXiv: 2406.11430, 2024, articleTitle=null, refAbstract=null), Reference(id=1327011181933474230, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhao Yanli, Gu A, Varma R, et al. PyTorch FSDP: Experiences on scaling fully sharded data parallel[J]. arXiv preprint, arXiv: 2304.11277, 2023, articleTitle=null, refAbstract=null), Reference(id=1327011181996388791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]//Proc of the 31st Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2017: 5998−6008, articleTitle=null, refAbstract=null), Reference(id=1327011182080274872, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Fu Yao. Challenges in deploying long-context transformers: A theoretical peak performance analysis[J]. arXiv preprint, arXiv: 2405. 08944, 2024, articleTitle=null, refAbstract=null), Reference(id=1327011182147383737, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Kwon W, Li Zhuohan, Zhuang Siyuan, et al. Efficient memory management for large language model serving with PagedAttention [C]//Proc of the 29th ACM SIGOPS Symp on Operating Systems Principles. New York: ACM, 2023: 611−626, articleTitle=null, refAbstract=null), Reference(id=1327011182264824250, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Yu G, Jeong J, Kim G, et al. ORCA: A distributed serving system for transformer-based generative models[C]//Proc of the 16th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2022: 521−538, articleTitle=null, refAbstract=null), Reference(id=1327011182373876155, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Lin Biao, Peng Ting, Zhang Chen, et al. Infinite-LLM: Efficient LLM service for long context with DistAttention and distributed KVCache[J]. arXiv preprint, arXiv: 2401. 02669, 2024, articleTitle=null, refAbstract=null), Reference(id=1327011182440985020, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Prabhu R, Nayak A, Mohan J, et al. vAttention: Dynamic memory management for serving LLMs without PagedAttention[J]. arXiv preprint, arXiv: 2405.04437, 2024, articleTitle=null, refAbstract=null), Reference(id=1327011182529065406, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Lin Ruihao, Peng Ting, Zhang Chen, et al. LightLLM: Efficient inference and serving framework for large language models[EB/OL]. [2024-07-18]. https://github.com/ModelTC/lightllm, articleTitle=null, refAbstract=null), Reference(id=1327011182659088831, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Zhenyu, Sheng Ying, Zhou Tianyi, et al. H2O: Heavy-hitter oracle for efficient generative inference of large language models[C]//Proc of the 37th Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2023: 1234−1245, articleTitle=null, refAbstract=null), Reference(id=1327011182780723648, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Tang Jiaming, Zhao Yilong, Zhu Kan, et al. Quest: Query-aware sparsity for efficient long-context LLM inference[C]//Proc of the 41st Int Conf on Machine Learning. New York: ICML, 2024: 123−134, articleTitle=null, refAbstract=null), Reference(id=1327011182940107201, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=An Chenxin, Gong Shanshan, Zhong Ming, et al. PyramidInfer: Efficient long-context inference for large language models[C]//Proc of the 38th Int Conf on Machine Learning. New York: ICML, 2024: 567−579, articleTitle=null, refAbstract=null), Reference(id=1327011183011410370, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Adnan M, Arunkumar A, Jain G, et al. Keyformer: KV cache reduction through key tokens selection for efficient generative inference[C/OL]//Proc of the 6th Conf on Machine Learning and Systems. 2024[2024-07-19]. https://proceedings.mlsys.org/paper_files/paper/2024/file/48fecef47b19fe501d27d338b6d52582-Paper-Conference.pdf, articleTitle=null, refAbstract=null), Reference(id=1327011183086907843, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Dao T, Fu D, Ermon S, et al. FlashAttention: Fast and memory-efficient exact attention with IO-awareness[C]//Proc of the 36th Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2022: 67−78, articleTitle=null, refAbstract=null), Reference(id=1327011183158211012, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Dao T, Ge Yuhao, Fu D, et al. FlashAttentio-2: Faster attention with better parallelism and work partitioning[J]. arXiv preprint, arXiv: 2307.08691, 2023, articleTitle=null, refAbstract=null), Reference(id=1327011183313400261, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Dao T, Haziza D, Massa F, Sizov G, et at. Flash-decoding for long-context inference [EB/OL]. 2023 [2024-01-27]. https://crfm.stanford. edu/2023/10/12/ffashdecoding. html, articleTitle=null, refAbstract=null), Reference(id=1327011183439229382, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Ke Hong, Dai Guohao, Xu Jiaming, et al. FlashDecoding++: Faster large language model inference on GPUs[C/OL]//Proc of the 7th Conf on Machine Learning and Systems. 2024[2024-07-18]. https://proceedings.mlsys.org/paper_files/paper/2024/file/5321b1dabcd2be188d796c21b733e8c7-Paper-Conference.pdf, articleTitle=null, refAbstract=null), Reference(id=1327011183548281287, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=HuggingFace. Transformers: State-of-the-art machine learning for pytorch, tensorflow, and JAX[EB/OL]. 2024[2024-07-18]. https://github.com/hugging face /transformers, articleTitle=null, refAbstract=null), Reference(id=1327011185226002889, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Aminabadi R, Rajbhandari S, Awan A, et al. Deepspeed-inference: Enabling efficient inference of transformer models at unprecedented scale[C/OL]//Proc of the 22nd Int Conf for High Performance Computing, Networking, Storage and Analysis. Piscataway, NJ: IEEE, 2022[2024-07-18]. https://dl.acm.org/doi/abs/10.5555/3571885.3571946, articleTitle=null, refAbstract=null), Reference(id=1327011185309888970, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=NVIDIA. Fastertransformer: About transformer related optimization, including BERT, GPT[EB/OL]. 2017[2024-07-18]. https://github.com/ NVIDIA/FasterTransformer, articleTitle=null, refAbstract=null), Reference(id=1327011185389580747, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Jin Yunho, Wu Chunfeng, Brooks D, et al. S3: Increasing GPU utilization during generative inference for higher throughput[C]//Proc of the 37th Int Conf on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc, 2024: 18015−18027, articleTitle=null, refAbstract=null), Reference(id=1327011185456689612, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=NVIDIA. NVLink & NVSwitch: Fastest HPC data center platform [EB/OL]. 2023[2024-07-18]. https://www.nvidia.com/en-sg/data-center/nvlink/, articleTitle=null, refAbstract=null), Reference(id=1327011185553158605, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Electronicshub. PCIe 5.0 vs PCIe 4.0: Which is best for high data transfer? [EB/OL]. 2023[2024-07-18]. https:// www .electronicshub.org/pcie-5- vs-pcie-4, articleTitle=null, refAbstract=null), Reference(id=1327011185628656078, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Github. ZeRO-Inference: Democratizing massive model inference DeepSpeed[EB/OL]. 2022[2024-07-15]. https://www.deepspeed.ai/2022/09/09/zero-inference.html, articleTitle=null, refAbstract=null), Reference(id=1327011185712542159, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Sheng Ying, Zheng Lianmin, Yuan Binhang, et al. FlexGen: High-throughput generative inference of large language models with a single GPU[C]//Proc of the 40th Int Conf on Machine Learning. New York: ICML, 2023: 31094−31116, articleTitle=null, refAbstract=null), Reference(id=1327011185779651024, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhao Xuanlei, Jia Bin, Zhou Haotian, et al. HeteGen: Efficient heterogeneous parallel inference for large language models on resource-constrained devices[C]//Proc of Machine Learning and Systems. New York: PMLR, 2024: 162−172, articleTitle=null, refAbstract=null), Reference(id=1327011185855148497, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Lei, Li Wei, Chen Jian, et al. Efficient and economic large language model inference with attention offloading[J]. arXiv preprint, arXiv: 2405.01814, 2024, articleTitle=null, refAbstract=null), Reference(id=1327011185943228882, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Bingyang, Zhong Yinmin, Zhang Zili, et al. Fast distributed inference serving for large language models[C]//Proc of the 40th Int Conf on Machine Learning. New York: PMLR, 2023: 1234−1245, articleTitle=null, refAbstract=null), Reference(id=1327011186014532051, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Lee W, Lee J, Seo J, et al. InfiniGen: Efficient generative inference of large language models with dynamic KV cache management[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2024: 567−579, articleTitle=null, refAbstract=null), Reference(id=1327011186077446612, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Song Yixin, Mi Zeyu, Xie Haotong, et al. PowerInfer: Fast large language model serving with a consumer-grade GPU[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2023: 123−134, articleTitle=null, refAbstract=null), Reference(id=1327011186144555477, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Alizadeh K, Mirzadeh I, Belenko D, et al. LLM in a flash: Efficient large language model inference with limited memory[C]//Proc of Int Conf on Machine Learning. New York: ICML, 2023: 123−134, articleTitle=null, refAbstract=null), Reference(id=1327011186203275734, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=42, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Microsoft. DeepSpeed[EB/OL]. 2023[2024-07-15]. https://github.com/microsoft/DeepSpeed, articleTitle=null, refAbstract=null), Reference(id=1327011186278773207, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=43, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Agrawal A, Panwar A, Mohan J, et al. Sarathi: Efficient LLM inference by piggybacking decodes with chunked prefills[J]. arXiv preprint, arXiv: 2308.16369, 2023, articleTitle=null, refAbstract=null), Reference(id=1327011186345882072, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=44, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Holmes C, Tanaka M, Wyatt M, et al. DeepSpeed-FastGen: High-throughput text generation for LLMs via MII and DeepSpeed-Inference[J]. arXiv preprint, arXiv: 2401.08671, 2024, articleTitle=null, refAbstract=null), Reference(id=1327011186412990937, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=45, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=Agrawal A, Kedia N, Panwar A, et al. Taming throughput-latency tradeoff in LLM inference with sarathi-serve[J]. arXiv preprint, arXiv: 2403.02310, 2024, articleTitle=null, refAbstract=null), Reference(id=1327011186475905498, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=46, rfOrder=45, authorNames=null, journalName=null, refType=null, unstructuredReference=Qin Ruoyu, Li Zheming, He Weiran, et al. Mooncake: A KVCache-centric disaggregated architecture for LLM serving[J]. arXiv preprint, arXiv: 2407.00079, 2024, articleTitle=null, refAbstract=null), Reference(id=1327011186639483355, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=47, rfOrder=46, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhong Yinmin, Liu Shengyu, Chen Junda, et al. DistServe: Disaggregating prefill and decoding for goodput-optimized large language model serving[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2024: 193−210, articleTitle=null, refAbstract=null), Reference(id=1327011186719175132, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=48, rfOrder=47, authorNames=null, journalName=null, refType=null, unstructuredReference=Hu Cunchen, Huang Heyang, Xu Liangliang, et al. Inference without interference: Disaggregate LLM inference for mixed downstream workloads[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2024: 193−210, articleTitle=null, refAbstract=null), Reference(id=1327011186790478301, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=49, rfOrder=48, authorNames=null, journalName=null, refType=null, unstructuredReference=Goiri Í, Maleki S, Bianchini R, et al. Splitwise: Efficient generative LLM inference using phase splitting[C]//Proc of the 28th ACM Symp on Operating Systems Principles. New York: ACM, 2024: 134−145, articleTitle=null, refAbstract=null), Reference(id=1327011186857587166, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=50, rfOrder=49, authorNames=null, journalName=null, refType=null, unstructuredReference=Strati F, McAllister S, Phanishayee A, et al. DéjàVu: KV-cache streaming for fast, fault-tolerant generative LLM serving[C]//Proc of the 41st Int Conf on Machine Learning. New York: PMLR, 2024: 456−467, articleTitle=null, refAbstract=null), Reference(id=1327011186949861855, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=51, rfOrder=50, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhong Yuchen, Sheng Guangming, Liu Juncheng, et al. Swift: Expedited failure recovery for large-scale DNN training[C]//Proc of the 28th ACM SIGPLAN Annual Symp on Principles and Practice of Parallel Programming. New York: ACM, 2023: 447−449, articleTitle=null, refAbstract=null), Reference(id=1327011187025359328, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=52, rfOrder=51, authorNames=null, journalName=null, refType=null, unstructuredReference=Jeong J, Baek S, Ahn J, et al. Fast and efficient model serving using multi-gpus with direct-host-access[C]//Proc of the 18th European Conf on Computer Systems. New York: ACM, 2023: 249−265, articleTitle=null, refAbstract=null), Reference(id=1327011187146994145, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=53, rfOrder=52, authorNames=null, journalName=null, refType=null, unstructuredReference=Guerraoui R, Schiper A. Fault-tolerance by replication in distributed systems[C]//Proc of the 1996 Ada-Europe Int Conf on Reliable Software Technologies. Berlin: Springer, 1996: 38−57, articleTitle=null, refAbstract=null), Reference(id=1327011187205714402, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=54, rfOrder=53, authorNames=null, journalName=null, refType=null, unstructuredReference=Kaashoek M, Tanenbaum A, Verstoep K, et al. On fault-tolerant data replication in distributed systems[C]//Proc of the 11th IEEE Int Conf on Distributed Computing Systems. Piscataway, NJ: IEEE, 1991: 222−230, articleTitle=null, refAbstract=null), Reference(id=1327011187264434659, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=55, rfOrder=54, authorNames=null, journalName=null, refType=null, unstructuredReference=Mohan J, Phanishayee A, Chidambaram V, et al. CheckFreq: Frequent, fine-grained DNN checkpointing[C]//Proc of the 19th USENIX Conf on File and Storage Technologies. Berkeley, CA: USENIX Association, 2021: 203−216, articleTitle=null, refAbstract=null), Reference(id=1327011187331543524, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=56, rfOrder=55, authorNames=null, journalName=null, refType=null, unstructuredReference=Amazon SageMaker. Machine learning service[EB/OL]. [2024-07-18]. https://aws.amazon.com/pm/sagemaker/, articleTitle=null, refAbstract=null), Reference(id=1327011187390263781, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=57, rfOrder=56, authorNames=null, journalName=null, refType=null, unstructuredReference=Microsoft. Azure ML[EB/OL]. [2024-07-18]. https://learn.microsoft.com/ en-us/ azure/machine-learning, articleTitle=null, refAbstract=null), Reference(id=1327011187453178342, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=58, rfOrder=57, authorNames=null, journalName=null, refType=null, unstructuredReference=The KServe. KServe[EB/OL]. [2024-07-18]. https:// github.com/kserve/ kserve, articleTitle=null, refAbstract=null), Reference(id=1327011187516092903, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=59, rfOrder=58, authorNames=null, journalName=null, refType=null, unstructuredReference=Fu Yao, Xue Leyang, Huang Yeqi, et al. ServerlessLLM: Locality-enhanced serverless inference for large language models[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2024: 123−134, articleTitle=null, refAbstract=null), Reference(id=1327011187612561896, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=60, rfOrder=59, authorNames=null, journalName=null, refType=null, unstructuredReference=PhoenixNAP. NVMe vs SATA: Storage comparison & differences [EB/OL]. [2024-07-18]. https://phoenixnap.com/kb/nvme-vs-sata, articleTitle=null, refAbstract=null), Reference(id=1327011187679670761, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=61, rfOrder=60, authorNames=null, journalName=null, refType=null, unstructuredReference=Enterprise Storage Forum. NVMe vs SATA: Comparing storage technologies[EB/OL]. [2024-07-15]. https://www.enterprisestorageforum. com/storage/nvme-vs-sata/, articleTitle=null, refAbstract=null), Reference(id=1327011187738391018, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=62, rfOrder=61, authorNames=null, journalName=null, refType=null, unstructuredReference=Amazon Web Services. Amazon simple storage service (S3)[EB/OL]. [2024-07-15]. https://aws.amazon.com/s3/, articleTitle=null, refAbstract=null), Reference(id=1327011187801305579, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=63, rfOrder=62, authorNames=null, journalName=null, refType=null, unstructuredReference=Ge Xuran, Lai Mingche, Liu Yang, et al. SpacKV: A pmem-aware key-value separation store based on LSM-tree[C]//Proc of the 19th Int Conf on Network and Parallel Computing. Berlin: Springer, 2022: 327−339, articleTitle=null, refAbstract=null), Reference(id=1327011187864220140, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=64, rfOrder=63, authorNames=null, journalName=null, refType=null, unstructuredReference=Lai Zhiquan, Li Shengwei, Tang Xudong, et al. Merak: An efficient distributed DNN training framework with automated 3D parallelism for giant foundation models[J]. IEEE Transactions on Parallel and Distributed Systems, 2023, 34(5): 1466−1478, articleTitle=null, refAbstract=null), Reference(id=1327011187939717613, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=65, rfOrder=64, authorNames=null, journalName=null, refType=null, unstructuredReference=Compute Express Link Consortium. Compute express link (CXL)[EB/ OL]. [2024-07-18]. https://www.computeexpresslink.org, articleTitle=null, refAbstract=null), Reference(id=1327011188023603694, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=66, rfOrder=65, authorNames=null, journalName=null, refType=null, unstructuredReference=ARM. Computational Storage: A new way to boost performance[EB/OL]. [2024-07-18]. https://newsroom.arm.com/blog/computational-storage, articleTitle=null, refAbstract=null), Reference(id=1327011189718102513, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=67, rfOrder=66, authorNames=null, journalName=null, refType=null, unstructuredReference=冯杨洋,汪庆,谢旻晖,等. 从BERT到ChatGPT:大语言模型训练中的存储系统挑战与技术发展[J]. 计算机研究与发展,2024,61(4):809−823, articleTitle=null, refAbstract=null), Reference(id=1327011189843931634, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=67, rfOrder=67, authorNames=null, journalName=null, refType=null, unstructuredReference=Feng Yangyang, Wang Qing, Xie Minhui, et al. From BERT to ChatGPT: Challenges and technical development of storage systems for large model training[J]. Journal of Computer Research and Development, 2024, 61(4): 809−823 (in Chinese), articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1327011169279258946, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=1, ext=[AuthorCompanyExt(id=1327011169304424771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1College of Computer Science and Technology, National University of Defense Technology, Changsha 410073), AuthorCompanyExt(id=1327011169337979204, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169279258946, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
1国防科技大学计算机学院 长沙 410073)]), AuthorCompany(id=1327011169434448197, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=2, ext=[AuthorCompanyExt(id=1327011169442836806, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169434448197, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
2School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou 510006), AuthorCompanyExt(id=1327011169451225415, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169434448197, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
2中山大学计算机学院 广州 510006)]), AuthorCompany(id=1327011169518334280, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, xref=3, ext=[AuthorCompanyExt(id=1327011169530917193, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169518334280, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
3School of Informatics, University of Edinburgh, Edinburgh EH89AB), AuthorCompanyExt(id=1327011169539305802, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, companyId=1327011169518334280, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=
3爱丁堡大学信息学院 爱丁堡 EH89AB)])], figs=[ArticleFig(id=1327011174069154175, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 1, caption=
Overview architecture of Transformer model, figureFileSmall=Z9LjR4WvNXw/+f7o6/pMKQ==, figureFileBig=hUlcZ8WpR7t7IbklUh6OTA==, tableContent=null), ArticleFig(id=1327011174132068736, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图1, caption=
Transformer模型整体架构图, figureFileSmall=Z9LjR4WvNXw/+f7o6/pMKQ==, figureFileBig=hUlcZ8WpR7t7IbklUh6OTA==, tableContent=null), ArticleFig(id=1327011174354366849, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 2, caption=
Illustration of ORCA internal fragmentation[16], figureFileSmall=vtSWQaOZ+XuSCGf/tucv9A==, figureFileBig=HFe5bPaCDgLvL/I22yM9Fg==, tableContent=null), ArticleFig(id=1327011174417281410, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图2, caption=
ORCA内部碎片示意图[16], figureFileSmall=vtSWQaOZ+XuSCGf/tucv9A==, figureFileBig=HFe5bPaCDgLvL/I22yM9Fg==, tableContent=null), ArticleFig(id=1327011174484390275, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 3, caption=
Overview architecture of vLLM [15], figureFileSmall=kash7qHj6NUmhWbcohFUyg==, figureFileBig=M8U2yLmZolDYmWT8nkK1Pg==, tableContent=null), ArticleFig(id=1327011174547304836, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图3, caption=
vLLM总体架构[15], figureFileSmall=kash7qHj6NUmhWbcohFUyg==, figureFileBig=M8U2yLmZolDYmWT8nkK1Pg==, tableContent=null), ArticleFig(id=1327011176220832133, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 4, caption=
Address translation from logical KV blocks in vLLM to physical memory blocks in OS[15], figureFileSmall=9Li+1QFBOI1UBeSgkF7REg==, figureFileBig=Tt4cdhwKtug6bEBnE0nkhg==, tableContent=null), ArticleFig(id=1327011176296329606, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图4, caption=
vLLM 逻辑键值块到操作系统物理内存块的地址转换[15], figureFileSmall=9Li+1QFBOI1UBeSgkF7REg==, figureFileBig=Tt4cdhwKtug6bEBnE0nkhg==, tableContent=null), ArticleFig(id=1327011176451518855, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 5, caption=
Architecture of the rManager design[17], figureFileSmall=Euyz46GuBhDSlOMFyG3/QA==, figureFileBig=KKN/j/ApJFrf83gjSc8DQg==, tableContent=null), ArticleFig(id=1327011176531210632, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图5, caption=
rManager设计架构[17], figureFileSmall=Euyz46GuBhDSlOMFyG3/QA==, figureFileBig=KKN/j/ApJFrf83gjSc8DQg==, tableContent=null), ArticleFig(id=1327011176610902409, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 6, caption=
Token level KV cache management of LightLLM[19], figureFileSmall=BjCMdGXFHbB3BWGbvOsa0Q==, figureFileBig=/Y7rzLVULdE7q3hp1hG0RQ==, tableContent=null), ArticleFig(id=1327011176703177098, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图6, caption=
LightLLM的标记级键值缓存管理[19], figureFileSmall=BjCMdGXFHbB3BWGbvOsa0Q==, figureFileBig=/Y7rzLVULdE7q3hp1hG0RQ==, tableContent=null), ArticleFig(id=1327011176820617611, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 7, caption=
Illustration of attention calculation, figureFileSmall=EBe9QgXuQYtVbKChXtJbqg==, figureFileBig=BufDFrNr60ES3CM1b/a5ZQ==, tableContent=null), ArticleFig(id=1327011176904503692, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图7, caption=
注意力计算示意图, figureFileSmall=EBe9QgXuQYtVbKChXtJbqg==, figureFileBig=BufDFrNr60ES3CM1b/a5ZQ==, tableContent=null), ArticleFig(id=1327011176984195469, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 8, caption=
Illustration of FlashAttention calculation[24], figureFileSmall=xVkRCFZhSpGjio/v/7Dhxg==, figureFileBig=WatoHpfts9TPTRQHEXSYow==, tableContent=null), ArticleFig(id=1327011177072275854, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图8, caption=
FlashAttention注意力计算示意图[24], figureFileSmall=xVkRCFZhSpGjio/v/7Dhxg==, figureFileBig=WatoHpfts9TPTRQHEXSYow==, tableContent=null), ArticleFig(id=1327011177185522063, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 9, caption=
Zig-zag block scheduling in FlexGen[35], figureFileSmall=8yWvmHG2EuBs+JWmkUC2oA==, figureFileBig=RWNjzbX0vfpgUsPraO3d1A==, tableContent=null), ArticleFig(id=1327011177256825232, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图9, caption=
FlexGen的Z字形调度[35], figureFileSmall=8yWvmHG2EuBs+JWmkUC2oA==, figureFileBig=RWNjzbX0vfpgUsPraO3d1A==, tableContent=null), ArticleFig(id=1327011177336517009, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 10, caption=
Illustration of heterogeneous parallel computing architecture in FlexGen attention module [35], figureFileSmall=puJAh5y83hCcFok/mpt3/w==, figureFileBig=Wx/ZKENNN+BhC0pnuVUq1w==, tableContent=null), ArticleFig(id=1327011177407820178, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图10, caption=
FlexGen注意力模块异构并行计算架构示意图[35], figureFileSmall=puJAh5y83hCcFok/mpt3/w==, figureFileBig=Wx/ZKENNN+BhC0pnuVUq1w==, tableContent=null), ArticleFig(id=1327011177483317651, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 11, caption=
Illustration of heterogeneous parallel computing architecture in HeteGen attention module [36], figureFileSmall=t++F3pCwZ43Jj6kVywk3Jw==, figureFileBig=rP0vlKK68o3dpsHJIx6l5w==, tableContent=null), ArticleFig(id=1327011177567203732, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图11, caption=
HeteGen注意力模块异构并行计算架构示意图[36], figureFileSmall=t++F3pCwZ43Jj6kVywk3Jw==, figureFileBig=rP0vlKK68o3dpsHJIx6l5w==, tableContent=null), ArticleFig(id=1327011177642701206, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 12, caption=
Overlapping key value cache offloading with intermediate result transmission[38], figureFileSmall=Zcmqbi+BmNS3/wywnh2kCw==, figureFileBig=5lS+CcKFtf6rB752mPlUqw==, tableContent=null), ArticleFig(id=1327011177709810071, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图12, caption=
键值缓存卸载与中间结果传递重叠[38], figureFileSmall=Zcmqbi+BmNS3/wywnh2kCw==, figureFileBig=5lS+CcKFtf6rB752mPlUqw==, tableContent=null), ArticleFig(id=1327011177789501848, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 13, caption=
Overview architecture of InfiniGen[39], figureFileSmall=rwON2UpTeDwXyGWMJnllzA==, figureFileBig=6tH0AxyDEUEMYeQ4+2LVlg==, tableContent=null), ArticleFig(id=1327011177894359449, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图13, caption=
InfiniGen的总体架构[39], figureFileSmall=rwON2UpTeDwXyGWMJnllzA==, figureFileBig=6tH0AxyDEUEMYeQ4+2LVlg==, tableContent=null), ArticleFig(id=1327011177986634138, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 14, caption=
Overview architecture of the PowerInfer[40], figureFileSmall=uEACKJeSRMpdBeZREtDszg==, figureFileBig=JixDmckQQpgOu8yAbC0OjA==, tableContent=null), ArticleFig(id=1327011178255069595, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图14, caption=
PowerInfer的总体架构[40], figureFileSmall=uEACKJeSRMpdBeZREtDszg==, figureFileBig=JixDmckQQpgOu8yAbC0OjA==, tableContent=null), ArticleFig(id=1327011178338955676, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 15, caption=
Workflow of Mooncake’s inference instances[46], figureFileSmall=8ndIzEMByQVpWD83UEAdMw==, figureFileBig=wJIS3JfX3lua7bVO3IWMSw==, tableContent=null), ArticleFig(id=1327011178414453149, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图15, caption=
Mooncake的推理实例工作流[46], figureFileSmall=8ndIzEMByQVpWD83UEAdMw==, figureFileBig=wJIS3JfX3lua7bVO3IWMSw==, tableContent=null), ArticleFig(id=1327011178578031006, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 16, caption=
Pipeline failure recovery for DéjàVu[50], figureFileSmall=CsuZJzgJ7vCDM8I2jwdFMw==, figureFileBig=DKCiRUExphSIAwh/8rV0Yg==, tableContent=null), ArticleFig(id=1327011178640945567, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图16, caption=
DéjàVu的流水线故障恢复[50], figureFileSmall=CsuZJzgJ7vCDM8I2jwdFMw==, figureFileBig=DKCiRUExphSIAwh/8rV0Yg==, tableContent=null), ArticleFig(id=1327011178716443041, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Figure 17, caption=
Overview architecture of SeverlessLLM[59], figureFileSmall=9wZRIJuNhJrW630BVwuRMA==, figureFileBig=qZYIvrLTBleYQ+XV/eqW4A==, tableContent=null), ArticleFig(id=1327011178791940514, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=图17, caption=
SeverlessLLM的总体架构[59], figureFileSmall=9wZRIJuNhJrW630BVwuRMA==, figureFileBig=qZYIvrLTBleYQ+XV/eqW4A==, tableContent=null), ArticleFig(id=1327011178859049379, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Table 1, caption=
Comparison of Large Language Model Inference Systems Using Different Key Value Cache Management Schemes, figureFileSmall=null, figureFileBig=null, tableContent=
| 系统 | 物理显存分配粒度 | 显存连续性 | 内部碎片 |
|---|
| vLLM[15] | 逻辑键值块 | 否 | 较少 |
| ORCA[16] | 最大请求序列 | 是 | 较多 |
| Infinite-LLM[17] | 逻辑键值块 | 否 | 较少 |
| vAttention[18] | 物理页 | 是 | 较少 |
| LightLLM[19] | 标记 | 是 | 极少 |
), ArticleFig(id=1327011178934546852, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=表1, caption=
采用不同的键值缓存管理方案的大语言模型推理系统对比
, figureFileSmall=null, figureFileBig=null, tableContent=
| 系统 | 物理显存分配粒度 | 显存连续性 | 内部碎片 |
|---|
| vLLM[15] | 逻辑键值块 | 否 | 较少 |
| ORCA[16] | 最大请求序列 | 是 | 较多 |
| Infinite-LLM[17] | 逻辑键值块 | 否 | 较少 |
| vAttention[18] | 物理页 | 是 | 较少 |
| LightLLM[19] | 标记 | 是 | 极少 |
), ArticleFig(id=1327011179165233573, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Table 2, caption=
Comparison of Large Language Model Inference Systems Based on Heterogeneous Storage Architecture, figureFileSmall=null, figureFileBig=null, tableContent=
| 大语言模型推理系统 | 存储介质 | 卸载数据类型 | 卸载粒度 | 优化策略 | 优化目标 |
|---|
| 注:1)不同卸载粒度的具体解释. 层粒度:以神经网络的层级为单位卸载数据;张量粒度:以张量为单位卸载数据;元素粒度:以张量中的元素为单位卸载数据(例如,将某张量中50%的元素分配到CPU内存);标记粒度:以标记为基本单位卸载其对应的键值缓存. 2)不同优化策略的具体解释. H:异构并行计算;A:异步数据传输;O:张量卸载策略;S:智能显存分配与预取. 3)不同优化目标的具体解释. M:减少显存占用;T:减少I/O传输量;L:隐藏I/O传输延迟;符号“|”表示两者兼有. |
| ZeRO-Inference[34] | 显存+DRAM+SSD | 模型权重 | 层粒度 | A|S | M|L |
| FlexGen[35] | 显存+DRAM+SSD | 模型权重|激活量|键值缓存 | 张量|元素粒度 | H|A|O | M|T|L |
| HeteGen[36] | 显存+DRAM | 模型权重|激活量|键值缓存 | 张量粒度 | H|A|O|S | M|T|L |
| Lamina[37] | 显存+DRAM | 键值缓存 | 标记粒度 | H|A|O|S | M|T|L |
| FastServe[38] | 显存+DRAM | 键值缓存 | 标记粒度 | A|O|S | M|T|L |
| InfiniGen[39] | 显存+DRAM | 键值缓存 | 标记粒度 | A|O|S | M|T|L |
| PowerInfer[40] | 显存+DRAM | 模型权重 | 张量粒度 | H|A|O|S | M|T|L |
| LLM in a flash[41] | 显存+DRAM+SSD | 模型权重|激活量 | 张量|标记粒度 | H|A|O|S | M|T|L |
), ArticleFig(id=1327011180742291879, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=表2, caption=
基于异构存储架构的大语言模型推理系统对比
, figureFileSmall=null, figureFileBig=null, tableContent=
| 大语言模型推理系统 | 存储介质 | 卸载数据类型 | 卸载粒度 | 优化策略 | 优化目标 |
|---|
| 注:1)不同卸载粒度的具体解释. 层粒度:以神经网络的层级为单位卸载数据;张量粒度:以张量为单位卸载数据;元素粒度:以张量中的元素为单位卸载数据(例如,将某张量中50%的元素分配到CPU内存);标记粒度:以标记为基本单位卸载其对应的键值缓存. 2)不同优化策略的具体解释. H:异构并行计算;A:异步数据传输;O:张量卸载策略;S:智能显存分配与预取. 3)不同优化目标的具体解释. M:减少显存占用;T:减少I/O传输量;L:隐藏I/O传输延迟;符号“|”表示两者兼有. |
| ZeRO-Inference[34] | 显存+DRAM+SSD | 模型权重 | 层粒度 | A|S | M|L |
| FlexGen[35] | 显存+DRAM+SSD | 模型权重|激活量|键值缓存 | 张量|元素粒度 | H|A|O | M|T|L |
| HeteGen[36] | 显存+DRAM | 模型权重|激活量|键值缓存 | 张量粒度 | H|A|O|S | M|T|L |
| Lamina[37] | 显存+DRAM | 键值缓存 | 标记粒度 | H|A|O|S | M|T|L |
| FastServe[38] | 显存+DRAM | 键值缓存 | 标记粒度 | A|O|S | M|T|L |
| InfiniGen[39] | 显存+DRAM | 键值缓存 | 标记粒度 | A|O|S | M|T|L |
| PowerInfer[40] | 显存+DRAM | 模型权重 | 张量粒度 | H|A|O|S | M|T|L |
| LLM in a flash[41] | 显存+DRAM+SSD | 模型权重|激活量 | 张量|标记粒度 | H|A|O|S | M|T|L |
), ArticleFig(id=1327011180826177961, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=EN, label=Table 3, caption=
Comparison of Different Distributed Large Language Model Inference Systems, figureFileSmall=null, figureFileBig=null, tableContent=
| 系统 | 批处理策略 | 并行策略 | 键值缓存传输 | 调度级别 |
|---|
| Sarathi[43] | 分块预填充与附带 | 流水线|张量并行 | 同步传输 | 迭代级调度 |
| DeepSpeed-FastGen[44] | 分块预填充与附带 | 流水线并行 | 异步传输 | 迭代级|标记级调度 |
| Mooncake[46] | 预填充与解码分离 | 流水线|张量并行 | 异步传输 | 请求级|迭代级调度 |
| DistServe[47] | 预填充与解码分离 | 流水线|张量并行 | 异步传输 | 请求级|迭代级调度 |
| TetriInfer[48] | 预填充与解码分离 | 流水线并行 | 异步传输 | 请求级|迭代级调度 |
| Splitwise[49] | 预填充与解码分离 | 流水线|张量并行 | 异步传输 | 请求级|迭代级调度 |
| DéjàVu[50] | 预填充与解码分离 | 流水线|张量并行 | 异步传输 | 请求级|迭代级调度 |
), ArticleFig(id=1327011180893286826, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1327011159850463496, language=CN, label=表3, caption=
不同分布式大语言模型推理系统的对比
, figureFileSmall=null, figureFileBig=null, tableContent=
| 系统 | 批处理策略 | 并行策略 | 键值缓存传输 | 调度级别 |
|---|
| Sarathi[43] | 分块预填充与附带 | 流水线|张量并行 | 同步传输 | 迭代级调度 |
| DeepSpeed-FastGen[44] | 分块预填充与附带 | 流水线并行 | 异步传输 | 迭代级|标记级调度 |
| Mooncake[46] | 预填充与解码分离 | 流水线|张量并行 | 异步传输 | 请求级|迭代级调度 |
| DistServe[47] | 预填充与解码分离 | 流水线|张量并行 | 异步传输 | 请求级|迭代级调度 |
| TetriInfer[48] | 预填充与解码分离 | 流水线并行 | 异步传输 | 请求级|迭代级调度 |
| Splitwise[49] | 预填充与解码分离 | 流水线|张量并行 | 异步传输 | 请求级|迭代级调度 |
| DéjàVu[50] | 预填充与解码分离 | 流水线|张量并行 | 异步传输 | 请求级|迭代级调度 |
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202440628, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202440628, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202440628, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202440628, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1740758400000, fullTextJson=null, articleText=null, reference=null)