Article(id=1325495615110656219, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325495515563041259, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202550376, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=research-article, receivedDate=1748966400000, receivedDateStr=2025-06-04, revisedDate=1763654400000, revisedDateStr=2025-11-21, acceptedDate=1764259200000, acceptedDateStr=2025-11-28, onlineDate=1793952381160, onlineDateStr=2026-11-06, pubDate=1772294400000, pubDateStr=2026-03-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1793952381160, onlineIssueDateStr=2026-11-06, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1793952381160, creator=13701087609, updateTime=1793952381160, updator=13701087609, issue=Issue{id=1325495515563041259, tenantId=1146029695717560320, journalId=1324759914146529478, year='2026', volume='63', issue='3', pageStart='539', pageEnd='823', issueExtLink='null', onlineDate='null', pubDate='1772294400000', pubDateStr='2026-03-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1793952357425, creator='13701087609', updateTime=1793952549954, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext={EN=IssueExt(id=1325496323193057998, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325495515563041259, language=EN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1325496323193057999, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1325495515563041259, language=CN, specialIssueTitle=, coverIllustrator=null, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null, downloadFileDto=null}, startPage=567, endPage=584, ext={EN=ArticleExt(id=1325495615563641052, articleId=1325495615110656219, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Storage Arrays Global Cache Technology Based on CXL, columnId=1325495518645859037, journalTitle=Journal of Computer Research and Development, columnName=Special Issue on Data Storage and Processing for New Quality Productive Forces, runingTitle=null, highlight=null, articleAbstract=

Global cache is a core technology for enterprise-grade storage arrays, whose primary function is to improve data write performance. I/O operations return immediately once data is written into the memory cache of storage controllers, without waiting for the data to be persisted to slow HDDs or SSDs. However, the design of global cache in enterprise-grade storage products must comprehensively consider factors such as the number of replicas, system high availability, and failure scenarios. Constrained by the mutual interactions of these factors, the number of controllers in storage clusters typically cannot exceed 4, and system high availability fails to support the scenario where three out of four controllers fail concurrently. Furthermore, the transmission of cache replica data between controllers via RDMA(remote direct memory access) technology also impairs storage performance. Thus, a memory-disaggregated global cache system prototype is designed based on the latest CXL(compute express link) technology. With memory cache disaggregated from within controllers to external, the system offers three key advantages: first, controllers and memory cache can scale out independently, supporting multiple cache replicas; second, controller failures do not cause cache data loss, supporting concurrent failures of n−1 out of n controllers; third, CXL technology significantly optimizes storage performance while enhancing system scalability and high availability. Experimental results demonstrate that the CXL-based global cache achieves a 38% performance improvement compared with the RDMA-based global cache.

, authors=Qiang Li1, Tao Liu1, Xueqing Zhang1, 2, *, Xianwei Meng3, Minggang Sun1, Bole Li3, Xi Chen1, authorsList=Qiang Li, Tao Liu, Xueqing Zhang, Xianwei Meng, Minggang Sun, Bole Li, Xi Chen, authorCompany=null, correspAuthors=Xueqing Zhang, authorNote=

born in 1982. PhD, professor. His main research interests include computer architecture, storage system, and memory pooling

born in 1983. Master, senior engineer. His main research interests include centralized storage architecture and hardware design

born in 1984. PhD candidate, assistant professor. His main research interests include storage system and system architecture

born in 1986. Master. His main research interests include storage system design and memory pooling

born in 1984. Bachelor, senior engineer. Member of CCF. His main research interests include storage software architecture and storage system performance

born in 1984. Master. His main research interests include storage technology and hardware design

born in 1993. Master. Member of CCF. His main research interests include server system architecture and hardware design

, correspAuthorsNote=null, copyrightStatement=Copyright © 2026 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=1325495624552034578, articleId=1325495615110656219, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=基于CXL的存储阵列全局缓存技术, columnId=1325495520386495199, journalTitle=计算机研究与发展, columnName=面向新质生产力的数据存储与处理专题, runingTitle=null, highlight=null, articleAbstract=

全局缓存是企业级存储阵列的关键技术,其核心功能是提高数据的写入速度:用户I/O写入存储控制器的内存缓存后立即返回,无需等待数据持久化至低速HDD或SSD设备。然而,企业级存储产品的全局缓存设计需综合考虑副本数量、系统高可用及故障场景等因素。受这些因素的相互制约,存储集群的控制器数量通常难以超过4个,且系统可用性无法支持4个控制器同时坏3个。此外,控制器间通过RDMA(remote direct memory access)传输缓存副本数据,也会影响存储性能。为此,基于最新的CXL(compute express link)技术,设计了内存分离式的全局缓存系统原型。内存缓存从控制器内部解耦至外部后,系统具有以下优势:控制器与内存缓存可独立横向扩展,可以支持多个缓存副本;控制器故障不会导致缓存数据丢失,n个控制器可以同时故障n−1个;CXL技术在提升系统扩展性与可用性的同时,能显著优化存储性能。实验结果表明,CXL全局缓存性能比RDMA全局缓存提升38%。

, authors=李强1, 刘涛1, 张雪庆1, 2, *, 孟宪伟3, 孙明刚1, 李博乐3, 陈曦1, authorsList=李强, 刘涛, 张雪庆, 孟宪伟, 孙明刚, 李博乐, 陈曦, authorCompany=null, correspAuthors=张雪庆, authorNote=

1982年生。博士,正高级工程师。主要研究方向为计算机体系结构、存储系统、内存池化

1983年生。硕士,高级工程师。主要研究方向为集中式存储架构、硬件设计

1984年生。博士研究生,助理研究员。主要研究方向为存储系统、系统架构

1986年生。硕士。主要研究方向为存储系统设计、内存池化

1984年生。学士,高级工程师。CCF会员。主要研究方向为存储软件架构、存储系统性能

1984年生。硕士。主要研究方向为存储技术、硬件设计

1993年生。硕士。CCF会员。主要研究方向为服务器系统架构、硬件设计

, correspAuthorsNote=
张雪庆()
, copyrightStatement=版权所有 © 2026 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=gdE1iNeYFbJJijVQC8N8Ww==, magXml=dm6CHKWM52ThWTEAJjxxEQ==, pdfUrl=null, pdf=3Qth99SKXWY1QOdMHUmbdQ==, pdfFileSize=2274793, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=EcqwN/iXnGuF6UAnIpOPyQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=3hUvPEb6XGvk0o+Izhcurg==, mapNumber=null, fund=null)}, authors=[Author(id=1325495625223123229, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=li.qiangbj@ieisystem.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325495625307009311, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625223123229, language=EN, stringName=Qiang Li, firstName=Qiang, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101, bio={"content":"

born in 1982. PhD, professor. His main research interests include computer architecture, storage system, and memory pooling

"}, bioImg=null, bioContent=

born in 1982. PhD, professor. His main research interests include computer architecture, storage system, and memory pooling

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495625390895392, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625223123229, language=CN, stringName=李强, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1浪潮电子信息产业股份有限公司 济南 250101, bio={"img":"Zo5PAPNV2Sv+zLmXuAiJTw==","content":"

1982年生。博士,正高级工程师。主要研究方向为计算机体系结构、存储系统、内存池化

"}, bioImg=Zo5PAPNV2Sv+zLmXuAiJTw==, bioContent=

1982年生。博士,正高级工程师。主要研究方向为计算机体系结构、存储系统、内存池化

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)])]), Author(id=1325495625546084642, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, orderNo=1, 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=1325495625646747940, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625546084642, language=EN, stringName=Tao Liu, firstName=Tao, middleName=null, lastName=Liu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101, bio={"content":"

born in 1983. Master, senior engineer. His main research interests include centralized storage architecture and hardware design

"}, bioImg=null, bioContent=

born in 1983. Master, senior engineer. His main research interests include centralized storage architecture and hardware design

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495625730634021, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625546084642, language=CN, stringName=刘涛, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1浪潮电子信息产业股份有限公司 济南 250101, bio={"img":"bRdcW4ZpWcmb+V3JzG8xMw==","content":"

1983年生。硕士,高级工程师。主要研究方向为集中式存储架构、硬件设计

"}, bioImg=bRdcW4ZpWcmb+V3JzG8xMw==, bioContent=

1983年生。硕士,高级工程师。主要研究方向为集中式存储架构、硬件设计

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)])]), Author(id=1325495625986486567, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zhangxq@ieisystem.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1325495626078761258, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625986486567, language=EN, stringName=Xueqing Zhang, firstName=Xueqing, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101
2School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074, bio={"content":"

born in 1984. PhD candidate, assistant professor. His main research interests include storage system and system architecture

"}, bioImg=null, bioContent=

born in 1984. PhD candidate, assistant professor. His main research interests include storage system and system architecture

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495626422694187, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625986486567, language=CN, stringName=张雪庆, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1浪潮电子信息产业股份有限公司 济南 250101
2华中科技大学计算机科学与技术学院 武汉 430074, bio={"img":"q/mOH6dy/SyvQfJq79akTA==","content":"

1984年生。博士研究生,助理研究员。主要研究方向为存储系统、系统架构

"}, bioImg=q/mOH6dy/SyvQfJq79akTA==, bioContent=

1984年生。博士研究生,助理研究员。主要研究方向为存储系统、系统架构

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)]), AuthorCompany(id=1325495625000825110, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=2, ext=[AuthorCompanyExt(id=1325495625013408023, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625000825110, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074), AuthorCompanyExt(id=1325495625021796632, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625000825110, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2华中科技大学计算机科学与技术学院 武汉 430074)])]), Author(id=1325495626540134701, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=1325495626812764463, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495626540134701, language=EN, stringName=Xianwei Meng, firstName=Xianwei, middleName=null, lastName=Meng, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095, bio={"content":"

born in 1986. Master. His main research interests include storage system design and memory pooling

"}, bioImg=null, bioContent=

born in 1986. Master. His main research interests include storage system design and memory pooling

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495626909233456, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495626540134701, language=CN, stringName=孟宪伟, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3浪潮(北京)电子信息产业有限公司 北京 100095, bio={"img":"ZVhoNEt/uPF/eD8JT51gYg==","content":"

1986年生。硕士。主要研究方向为存储系统设计、内存池化

"}, bioImg=ZVhoNEt/uPF/eD8JT51gYg==, bioContent=

1986年生。硕士。主要研究方向为存储系统设计、内存池化

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495625109877017, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=3, ext=[AuthorCompanyExt(id=1325495625122459930, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095), AuthorCompanyExt(id=1325495625135042843, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3浪潮(北京)电子信息产业有限公司 北京 100095)])]), Author(id=1325495627332858162, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=1325495627458687284, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495627332858162, language=EN, stringName=Minggang Sun, firstName=Minggang, middleName=null, lastName=Sun, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101, bio={"content":"

born in 1984. Bachelor, senior engineer. Member of CCF. His main research interests include storage software architecture and storage system performance

"}, bioImg=null, bioContent=

born in 1984. Bachelor, senior engineer. Member of CCF. His main research interests include storage software architecture and storage system performance

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495627706151221, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495627332858162, language=CN, stringName=孙明刚, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1浪潮电子信息产业股份有限公司 济南 250101, bio={"img":"4uNp/iUQuMLZ4KmcS6MLgw==","content":"

1984年生。学士,高级工程师。CCF会员。主要研究方向为存储软件架构、存储系统性能

"}, bioImg=4uNp/iUQuMLZ4KmcS6MLgw==, bioContent=

1984年生。学士,高级工程师。CCF会员。主要研究方向为存储软件架构、存储系统性能

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)])]), Author(id=1325495627815203127, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=1325495627915866425, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495627815203127, language=EN, stringName=Bole Li, firstName=Bole, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095, bio={"content":"

born in 1984. Master. His main research interests include storage technology and hardware design

"}, bioImg=null, bioContent=

born in 1984. Master. His main research interests include storage technology and hardware design

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495628062667066, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495627815203127, language=CN, stringName=李博乐, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3浪潮(北京)电子信息产业有限公司 北京 100095, bio={"img":"q1xKw1CBv+lr2AyQqSnCvg==","content":"

1984年生。硕士。主要研究方向为存储技术、硬件设计

"}, bioImg=q1xKw1CBv+lr2AyQqSnCvg==, bioContent=

1984年生。硕士。主要研究方向为存储技术、硬件设计

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495625109877017, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=3, ext=[AuthorCompanyExt(id=1325495625122459930, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095), AuthorCompanyExt(id=1325495625135042843, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3浪潮(北京)电子信息产业有限公司 北京 100095)])]), Author(id=1325495628133970236, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=1325495628222050622, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495628133970236, language=EN, stringName=Xi Chen, firstName=Xi, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101, bio={"content":"

born in 1993. Master. Member of CCF. His main research interests include server system architecture and hardware design

"}, bioImg=null, bioContent=

born in 1993. Master. Member of CCF. His main research interests include server system architecture and hardware design

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495628477903167, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495628133970236, language=CN, stringName=陈曦, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1浪潮电子信息产业股份有限公司 济南 250101, bio={"img":"lVlFZvsrLedDE2Ipq5l1rQ==","content":"

1993年生。硕士。CCF会员。主要研究方向为服务器系统架构、硬件设计

"}, bioImg=lVlFZvsrLedDE2Ipq5l1rQ==, bioContent=

1993年生。硕士。CCF会员。主要研究方向为服务器系统架构、硬件设计

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)])])], keywords=[Keyword(id=1325495628624703808, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, orderNo=1, keyword=enterprise storage), Keyword(id=1325495628930888001, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, orderNo=2, keyword=storage array), Keyword(id=1325495629031551298, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, orderNo=3, keyword=CXL), Keyword(id=1325495629316763971, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, orderNo=4, keyword=global cache), Keyword(id=1325495629430010180, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, orderNo=5, keyword=disaggregate architecture), Keyword(id=1325495629497119045, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, orderNo=1, keyword=企业存储), Keyword(id=1325495629820080454, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, orderNo=2, keyword=存储阵列), Keyword(id=1325495629937520967, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, orderNo=3, keyword=CXL), Keyword(id=1325495630243705160, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, orderNo=4, keyword=全局缓存), Keyword(id=1325495630356951369, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, orderNo=5, keyword=分离式架构)], refs=[Reference(id=1325495638451958135, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=CXL. CXL® specification - compute express[EB/OL]. [2025-05-24]. https://computeexpresslink.org/cxl-specification, articleTitle=null, refAbstract=null), Reference(id=1325495638548427128, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=舒继武, 数据存储架构与技术[M]. 北京: 人民邮电出版社, 2023: 50−58, articleTitle=null, refAbstract=null), Reference(id=1325495638644896121, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Shu Jiwu. Data Storage Architecture and Technology[M]. Beijing: Posts & Telecom Press, 2023: 50−58 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495638728782202, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=吴晨涛, 信息存储与IT管理[M]. 北京: 人民邮电出版社, 2015, articleTitle=null, refAbstract=null), Reference(id=1325495638808473979, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Wu Chentao. Information Storage and IT Management[M]. Beijing: Posts & Telecom Press, 2015 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495638879777148, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=浪潮信息. 浪潮信息高端存储系统 HF18000G6-CGF[EB/OL]. [2025-05-24]. https://www.ieisystem.com/product/storage/15598.html, articleTitle=null, refAbstract=null), Reference(id=1325495638972051837, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=IEIT. IEIT high-end storage system HF18000G6-CGF[EB/OL]. [2025-05-24]. https://www.ieisystem.com/product/storage/15598.html (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495639060132222, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=华为. OceanStor Dorado V6 高端全闪存存储产品技术白皮书[EB/OL]. [2025-05-24]. https: //e.huawei.com/cn/material/storage/all-flash-storage/42b9aab381c743b1b68c73db12931e3d, articleTitle=null, refAbstract=null), Reference(id=1325495639144018303, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Huawei. OceanStor Dorado V6 high-end all-flash storage product technical white paper[EB/OL]. [2025-05-24]. https://e.huawei.com/cn/material/storage/all-flash-storage/42b9aab381c743b1b68c73db12931e3d (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495639236292992, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=EMC. A detailed overview of the next generation PowerMax platform[EB/OL]. [2025-05-24]. https://infohub.delltechnologies.com/static/media/client/7phukh/DAM_3ddff5ea-f554-4502-9304-59a014d9daea.pdf, articleTitle=null, refAbstract=null), Reference(id=1325495639307596161, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=VastData. White paper: The VAST data platform[EB/OL]. [2025-05-24]. https://www.vastdata.com/whitepaper, articleTitle=null, refAbstract=null), Reference(id=1325495639404065154, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=HPE. HPE Alletra storage MP B10000 architecture[EB/OL]. [2025-05-24]. https://www.hpe.com/psnow/doc/a50008302enw?jumpId=in_psnow_ee8f0a3a-72f8-47de-bf38-7b94e0de4c77_gaiw, articleTitle=null, refAbstract=null), Reference(id=1325495639538282883, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Montage. Montage technology delivers the world’s first CXL™ memory eXpander controller[EB/OL]. [2025-05-24]. https://www.montage-tech.com/Press_Releases/20220506, articleTitle=null, refAbstract=null), Reference(id=1325495639626363268, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Roland K. Montage technology CXL add in card[EB/OL]. [2025-05-24]. https://drive.google.com/file/d/1bpVLz533F2YTVqN3mIqtr--ngiR35lBe/view?usp=share_link, articleTitle=null, refAbstract=null), Reference(id=1325495639810912645, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Samsung. Samsung electronics introduces industry’s first 512GB CXL memory module[EB/OL]. (2022-05-10)[2025-05-24]. https://semiconductor.samsung.com/us/news-events/news/samsung-electronics-introduces-industrys-first-512gb-cxl-memory-module/, articleTitle=null, refAbstract=null), Reference(id=1325495639907381638, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Samsung. CMM-B: Memory pooling appliance for rack computing[EB/OL]. (2022-05-10)[2025-05-24]. https://semiconductor.samsung.com/us/about-us/us-office/us-r-and-d-labs/memory-labs/cxl-memory-module-box/, articleTitle=null, refAbstract=null), Reference(id=1325495641597686151, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Gao Xianyang, Chen Xi, Wang Yun. Investigating the memory extension and disaggregated memory pooling system[C]//Proc of the 4th Int Conf on Electronic Information Engineering and Computer Science. Piscataway, NJ: IEEE, 2024: 667−671, articleTitle=null, refAbstract=null), Reference(id=1325495641681572232, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=XConn. XConn technologies launches the industry’s first and only hybrid CXL 2.0 and PCIe gen 5 switch[EB/OL]. (2023-08-08)[2025-05-24]. https://www.prnewswire.com/news-releases/xconn-technologies-launches-the-industrys-first-and-only-hybrid-cxl-2-0-and-pcie-gen-5-switch-301895313.html?tc=eml_cleartime, articleTitle=null, refAbstract=null), Reference(id=1325495641769652617, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Samsung. Samsung memory-semantic CXL SSD[EB/OL]. (2022-08-18)[2025-05-24]. https://www.servethehome.com/samsung-memory-semantic-cxl-ssd-at-fms-2022-powered-by-amd-xilinx/, articleTitle=null, refAbstract=null), Reference(id=1325495641874510218, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Jung M. Hello bytes, bye blocks: PCIe storage meets compute express link for memory expansion (CXL-SSD)[C]//Proc of the 14th ACM Workshop on Hot Topics in Storage and File Systems. New York: ACM, 2022: 45−51, articleTitle=null, refAbstract=null), Reference(id=1325495641970979211, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Memverge. Memory pooling and sharing solution[EB/OL]. (2023-08-08)[2025-05-24]. https://memverge.com/samsung-memverge-h3-platform-and-xconn-demonstrate-memory-pooling-and-sharing-for-endless-memory/, articleTitle=null, refAbstract=null), Reference(id=1325495642054865292, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Li Huaicheng, Berger D S, Hsu L, et al. Pond: CXL-based memory pooling systems for cloud platforms[C]//Proc of the 28th ACM Int Conf on Architectural Support for Programming Languages and Operating Systems. New York: ACM, 2023: 574−587, articleTitle=null, refAbstract=null), Reference(id=1325495642155528589, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Berger D S, Ernst D, Li H, et al. Design tradeoffs in CXL-based memory pools for public cloud platforms[J]. IEEE Micro, 2023, 43(2): 30−38, articleTitle=null, refAbstract=null), Reference(id=1325495642239414670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Gouk D, Lee S, Kwon M, et al. Direct access, high-performance memory disaggregation with directCXL[C]//Proc of USENIX Annual Technical Conf 2022. Berkeley, CA: USENIX Association, 2022: 287−294, articleTitle=null, refAbstract=null), Reference(id=1325495642365243791, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Sharma D D. Compute express link (CXL): Enabling heterogeneous data-centric computing with heterogeneous memory hierarchy[J]. IEEE Micro, 2022, 43(2): 99−109, articleTitle=null, refAbstract=null), Reference(id=1325495642478490000, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Gouk D, Kwon M, Bae H, et al. Memory pooling with CXL[J]. IEEE Micro, 2023, 43(2): 48−57, articleTitle=null, refAbstract=null), Reference(id=1325495642566570385, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Jinshu, Hadian H, Xu Hanchen, et al. Systematic CXL memory characterization and performance analysis at scale[C]// Proc of the 30th ACM Int Conf on Architectural Support for Programming Languages and Operating Systems. New York: ACM, 2025: 1203−1217, articleTitle=null, refAbstract=null), Reference(id=1325495642663039378, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Xi, Liu Jie, Wu Jianbo, et al. Exploring and evaluating real-world CXL: Use cases and system adoption[J]. arXiv preprint, arXiv: 2405.14209, 2024, articleTitle=null, refAbstract=null), Reference(id=1325495642751119763, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Sun Yan, Yuan Yifan, Yu Zeduo, et al. Demystifying CXL memory with genuine CXL-ready systems and devices[C]//Proc of the 56th Annual IEEE/ACM Int Symp on Microarchitecture. New York: ACM, 2023: 105−121, articleTitle=null, refAbstract=null), Reference(id=1325495642843394452, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Kwon M, Lee S, Jung M. Cache in hand: Expander-driven CXL prefetcher for next generation CXL-SSD[C]//Proc of the 15th ACM Workshop on Hot Topics in Storage and File Systems. New York: ACM, 2023: 24−30, articleTitle=null, refAbstract=null), Reference(id=1325495642948252053, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Shaopeng, Kim M, Nam S, et al. Overcoming the memory wall with CXL-enabled SSDs[C]//Proc of USENIX Annual Technical Conf 2023. Berkeley, CA: USENIX Association, 2023: 601−617, articleTitle=null, refAbstract=null), Reference(id=1325495643019555222, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Zhonghua, Guo Yixing, Lu Kai, et al. Rcmp: Reconstructing RDMA-based memory disaggregation via CXL[J]. ACM Transactions on Architecture and Code Optimization, 2024, 21(1): 1−26, articleTitle=null, refAbstract=null), Reference(id=1325495643103441303, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Chenjiu, He Ke, Fan Ruiqi, et al. CXL over ethernet: A novel FPGA-based memory disaggregation design in data centers[C]//Proc of the 31st IEEE Annual Int Symp on Field-Programmable Custom Computing Machines. Piscataway, NJ: IEEE, 2023: 75−82, articleTitle=null, refAbstract=null), Reference(id=1325495643174744472, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Kim K, Kim H, So J, et al. SMT: Software-defined memory tiering for heterogeneous computing systems with CXL memory expander[J]. IEEE Micro, 2023, 43(2): 20−29, articleTitle=null, refAbstract=null), Reference(id=1325495643258630553, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Maruf H A, Wang Hao, Dhanotia A, et al. Tpp: Transparent page placement for CXL-enabled tiered-memory[C]//Proc of the 28th ACM Int Conf on Architectural Support for Programming Languages and Operating Systems. New York: ACM, 2023: 742−755, articleTitle=null, refAbstract=null), Reference(id=1325495643329933722, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Xiang Lingfeng, Lin Zhen, Deng Weishu, et al. Nomad: Non-exclusive memory tiering via transactional page migration[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2024: 19−35, articleTitle=null, refAbstract=null), Reference(id=1325495643409625499, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhong Yuhong, Berger D S, Waldspurger C, et al. Managing memory tiers with CXL in virtualized environments[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2024: 37−56, articleTitle=null, refAbstract=null), Reference(id=1325495643493511580, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Hancheng, Dai Haipeng, Chen S, et al. Rethinking Hash tables: Challenges and opportunities with compute express link (CXL)[C]//Proc of the ACM Turing Award Celebration Conf-China 2024. New York: ACM, 2024: 23−27, articleTitle=null, refAbstract=null), Reference(id=1325495643598369181, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Huang Wentao, Sha Mo, Lu Mian, et al. Bandwidth expansion via CXL: A pathway to accelerating in-memory analytical processing[C]//Proc of the 16th Int Workshop on Accelerating Analytics and Data Management Systems Using Modern Processor and Storage Architectures. New York: ACM, 2024: 1−12, articleTitle=null, refAbstract=null), Reference(id=1325495643669672350, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Ahn M, Willhalm T, May N, et al. An examination of CXL memory use cases for in-memory database management systems using SAP HANA[J]. Proceedings of the VLDB Endowment, 2024, 17(12): 3827−3840, articleTitle=null, refAbstract=null), Reference(id=1325495643766141343, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Gouk D, Kang S, Bae H, et al. Breaking barriers: Expanding GPU memory with sub-two digit nanosecond latency CXL controller//Proc of the 16th ACM Workshop on Hot Topics in Storage and File Systems. New York: ACM, 2024: 108−115, articleTitle=null, refAbstract=null), Reference(id=1325495643845833120, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=41, authorNames=null, journalName=null, refType=null, unstructuredReference=Park S S, Kim K S, So J, et al. An LPDDR-based CXL-PNM platform for TCO-efficient inference of transformer-based large language models[C]//Proc of the 30th IEEE Int Symp on High-Performance Computer Architecture. Piscataway, NJ: IEEE, 2024: 970−982, articleTitle=null, refAbstract=null), Reference(id=1325495643917136289, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=42, authorNames=null, journalName=null, refType=null, unstructuredReference=Liu Haifeng, Zheng Long, Huang Yu, et al. Enabling efficient large recommendation model training with near CXL memory processing[C]//Proc of the 51st ACM/IEEE Annual Int Symp on Computer Architecture. Piscataway, NJ: IEEE, 2024: 382−395, articleTitle=null, refAbstract=null), Reference(id=1325495643992633762, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=43, authorNames=null, journalName=null, refType=null, unstructuredReference=Arif M, Assogba K, Rafique M M, et al. Exploiting CXL-based memory for distributed deep learning[C]//Proc of the 51st Int Conf on Parallel Processing. Piscataway, NJ: IEEE, 2022: 1−11, articleTitle=null, refAbstract=null), Reference(id=1325495644063936931, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=44, authorNames=null, journalName=null, refType=null, unstructuredReference=Jang J, Choi H, Bae H, et al. CXL-ANNS: Software-hardware collaborative memory disaggregation and computation for billion-scale approximate nearest neighbor search[C]//Proc of USENIX Annual Technical Conf 2023. Berkeley, CA: USENIX Association, 2023: 585−600, articleTitle=null, refAbstract=null), Reference(id=1325495644143628708, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=42, rfOrder=45, authorNames=null, journalName=null, refType=null, unstructuredReference=Ma Teng, Liu Zheng, Wei Chengkun, et al. HydraRPC: RPC in the CXL era[C]//Proc of USENIX Annual Technical Conf 2024. Berkeley, CA: USENIX Association, 2024: 387−395, articleTitle=null, refAbstract=null), Reference(id=1325495644244292005, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=43, rfOrder=46, authorNames=null, journalName=null, refType=null, unstructuredReference=Reichard G, Luis R R. MTRR(memory type range register) control[EB/OL]. [2025-05-24]. https://www.kernel.org/doc/html/next/x86/mtrr.html, articleTitle=null, refAbstract=null), Reference(id=1325495644416258470, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=44, rfOrder=47, authorNames=null, journalName=null, refType=null, unstructuredReference=Memverge. Gismo: World’s first CXL-based multi-server shared memory[EB/OL]. (2023-05-23)[2025-05-24]. https://memverge.com/memverge-unveils-worlds-first-CXL-based-multi-server-shared-memory-at-isc/, articleTitle=null, refAbstract=null), Reference(id=1325495646077202855, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=45, rfOrder=48, authorNames=null, journalName=null, refType=null, unstructuredReference=Storage Performance Council. Storage performance council[EB/OL]. [2025-05-24]. https://www.storageperformance.org, articleTitle=null, refAbstract=null), Reference(id=1325495646165283240, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=46, rfOrder=49, authorNames=null, journalName=null, refType=null, unstructuredReference=舒继武, 新型存算分离架构技术展望[J]. 中国计算机学会通讯, 2022, 18(11): 53−60, articleTitle=null, refAbstract=null), Reference(id=1325495646261752233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=46, rfOrder=50, authorNames=null, journalName=null, refType=null, unstructuredReference=Shu Jiwu. Technical prospects of new compute-storage disaggregation architecture[J]. CCF Communications, 2022, 18(11): 53−60 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495646358221226, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=47, rfOrder=51, authorNames=null, journalName=null, refType=null, unstructuredReference=舒继武, 陈游旻, 汪庆, 等. 分离式数据中心的存储系统研究进展[J]. 中国科学: 信息科学, 2023, 53(8): 1503−1528, articleTitle=null, refAbstract=null), Reference(id=1325495646496633259, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=47, rfOrder=52, authorNames=null, journalName=null, refType=null, unstructuredReference=Shu Jiwu, Chen Youmin, Wang Qing, et al. Progress on storage systems for disaggregated data centers[J]. SCIENTIA SINICA Informationis, 2023, 53(8): 1503−1528 (in Chinese), articleTitle=null, refAbstract=null), Reference(id=1325495646727319980, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=48, rfOrder=53, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Jing, Li Chao, Wang Taolei, et al. Survey of disaggregated memory: Cross-layer technique insights for next-generation datacenters[J]. arXiv preprint, arXiv: 2503.20275, 2025, articleTitle=null, refAbstract=null), Reference(id=1325495646802817453, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=49, rfOrder=54, authorNames=null, journalName=null, refType=null, unstructuredReference=Yang Xinjun, Zhang Yingqiang, Chen Hao, et al. Unlocking the potential of CXL for disaggregated memory in cloud-native databases[C]//Proc of the 2025 Companion of Int Conf on Management of Data. New York: ACM, 2025: 689−702, articleTitle=null, refAbstract=null), Reference(id=1325495646903480750, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=50, rfOrder=55, authorNames=null, journalName=null, refType=null, unstructuredReference=Intel. Intel memory latency checker (MLC)[EB/OL]. [2025-05-24]. https://www.intel.cn/content/www/cn/zh/download/736633/736634/intel-memory-latency-checker-intel-mlc.html, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)]), AuthorCompany(id=1325495625000825110, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=2, ext=[AuthorCompanyExt(id=1325495625013408023, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625000825110, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074), AuthorCompanyExt(id=1325495625021796632, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625000825110, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2华中科技大学计算机科学与技术学院 武汉 430074)]), AuthorCompany(id=1325495625109877017, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=3, ext=[AuthorCompanyExt(id=1325495625122459930, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095), AuthorCompanyExt(id=1325495625135042843, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3浪潮(北京)电子信息产业有限公司 北京 100095)])], figs=[ArticleFig(id=1325495630642164042, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 1, caption=Four controller global cache(loop mirror), figureFileSmall=swDqKFab9xlfj1K94fBUQw==, figureFileBig=EcqwN/iXnGuF6UAnIpOPyQ==, tableContent=null), ArticleFig(id=1325495630709272907, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图1, caption=4控全局缓存(循环镜像), figureFileSmall=swDqKFab9xlfj1K94fBUQw==, figureFileBig=EcqwN/iXnGuF6UAnIpOPyQ==, tableContent=null), ArticleFig(id=1325495632726733132, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 2, caption=EMC global cache diagram, figureFileSmall=Zg0yP7euZ5rrjMaro/1vrQ==, figureFileBig=3fmqv+3a+86MnveUSR/78w==, tableContent=null), ArticleFig(id=1325495632827396429, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图2, caption=EMC全局缓存图, figureFileSmall=Zg0yP7euZ5rrjMaro/1vrQ==, figureFileBig=3fmqv+3a+86MnveUSR/78w==, tableContent=null), ArticleFig(id=1325495632902893902, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 3, caption=Global cache based on SCM, figureFileSmall=ussw6nTySjFG7pFIg1z7Gw==, figureFileBig=5R9F2mJj0rdwiBnE/MEo2Q==, tableContent=null), ArticleFig(id=1325495632999362895, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图3, caption=基于SCM的全局缓存, figureFileSmall=ussw6nTySjFG7pFIg1z7Gw==, figureFileBig=5R9F2mJj0rdwiBnE/MEo2Q==, tableContent=null), ArticleFig(id=1325495633066471760, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 4, caption=Architecture diagram of CXL global cache, figureFileSmall=Rass6ZuppcJhhZuVtUP5zQ==, figureFileBig=eo68um3agRqrqRiDefMxgQ==, tableContent=null), ArticleFig(id=1325495633183912273, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图4, caption=CXL全局缓存架构图, figureFileSmall=Rass6ZuppcJhhZuVtUP5zQ==, figureFileBig=eo68um3agRqrqRiDefMxgQ==, tableContent=null), ArticleFig(id=1325495633263604050, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 5, caption=Appearance diagram of JBOM hardware, figureFileSmall=Jv9+ZdCucWx86jR3NZeqNA==, figureFileBig=JsWH/e6bDbLYSRKV2Gv3YQ==, tableContent=null), ArticleFig(id=1325495633368461651, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图5, caption=JBOM硬件外观图, figureFileSmall=Jv9+ZdCucWx86jR3NZeqNA==, figureFileBig=JsWH/e6bDbLYSRKV2Gv3YQ==, tableContent=null), ArticleFig(id=1325495633456542036, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 6, caption=Topology diagram of JBOM board, figureFileSmall=m6iGmcjzUeyjjyzjbefgcA==, figureFileBig=qA1pQh99QjoMJeyLijUEoA==, tableContent=null), ArticleFig(id=1325495633578176853, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图6, caption=JBOM主板拓扑结构图, figureFileSmall=m6iGmcjzUeyjjyzjbefgcA==, figureFileBig=qA1pQh99QjoMJeyLijUEoA==, tableContent=null), ArticleFig(id=1325495633662062934, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 7, caption=Retimer card in OCP form factor, figureFileSmall=81ntisQWpoCIFS1BVwwcug==, figureFileBig=X/Uc0wN4OolxbMlgxwfgeQ==, tableContent=null), ArticleFig(id=1325495633745949015, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图7, caption=OCP形态的Retimer卡, figureFileSmall=81ntisQWpoCIFS1BVwwcug==, figureFileBig=X/Uc0wN4OolxbMlgxwfgeQ==, tableContent=null), ArticleFig(id=1325495633821446488, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 8, caption=Memory management module, figureFileSmall=Th/OmRIy+An+rFZP1UB7Og==, figureFileBig=W0QbHmZw5x7nzQrsb45RBg==, tableContent=null), ArticleFig(id=1325495633913721177, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图8, caption=内存管理模块, figureFileSmall=Th/OmRIy+An+rFZP1UB7Og==, figureFileBig=W0QbHmZw5x7nzQrsb45RBg==, tableContent=null), ArticleFig(id=1325495634026967386, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 9, caption=Design of two-level memory page table, figureFileSmall=I+8qRJ9kM5E94OULo47nsw==, figureFileBig=s2D002Wq7lYdm1YUCLa65g==, tableContent=null), ArticleFig(id=1325495634123436379, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图9, caption=2级内存页表设计, figureFileSmall=I+8qRJ9kM5E94OULo47nsw==, figureFileBig=s2D002Wq7lYdm1YUCLa65g==, tableContent=null), ArticleFig(id=1325495634219905372, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 10, caption=Three-level memory pool, figureFileSmall=H5T5mfGjf6HBXtbqUZyMKQ==, figureFileBig=5sof6YNLDNGr68FON5BARA==, tableContent=null), ArticleFig(id=1325495634303791453, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图10, caption=3级内存池, figureFileSmall=H5T5mfGjf6HBXtbqUZyMKQ==, figureFileBig=5sof6YNLDNGr68FON5BARA==, tableContent=null), ArticleFig(id=1325495634421231966, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 11, caption=I/O flowchart of CXL global cache, figureFileSmall=DdfJuVuhYefk+hK5pyPJtw==, figureFileBig=b2Z5l6G51bwx2ppAWJPejQ==, tableContent=null), ArticleFig(id=1325495634517700959, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图11, caption=CXL全局缓存的I/O流程图, figureFileSmall=DdfJuVuhYefk+hK5pyPJtw==, figureFileBig=b2Z5l6G51bwx2ppAWJPejQ==, tableContent=null), ArticleFig(id=1325495634593198432, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 12, caption=I/O flowchart of RDMA cache mirroring, figureFileSmall=846e+ZDKk884TI7eUyPHJg==, figureFileBig=LqvE3+/ehhwLgBKC8O5cnA==, tableContent=null), ArticleFig(id=1325495634677084513, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图12, caption=RDMA缓存镜像的I/O流程图, figureFileSmall=846e+ZDKk884TI7eUyPHJg==, figureFileBig=LqvE3+/ehhwLgBKC8O5cnA==, tableContent=null), ArticleFig(id=1325495634781942114, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 13, caption=Fault topology diagram of four controller cache mirroring, figureFileSmall=2fMl4lmv9Wm7ULpgDk0Njw==, figureFileBig=gCvgeZhRLS1801BIMc5B1w==, tableContent=null), ArticleFig(id=1325495634865828195, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图13, caption=4控缓存镜像的故障拓扑图, figureFileSmall=2fMl4lmv9Wm7ULpgDk0Njw==, figureFileBig=gCvgeZhRLS1801BIMc5B1w==, tableContent=null), ArticleFig(id=1325495634979074404, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 14, caption=Fault topology diagram of CXL global cache mirroring, figureFileSmall=i1E6Jm8izxzrcKIp4LcElA==, figureFileBig=O+n5IugmpP9HKyRkkyUmBA==, tableContent=null), ArticleFig(id=1325495635192983909, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图14, caption=CXL全局缓存镜像的故障拓扑图, figureFileSmall=i1E6Jm8izxzrcKIp4LcElA==, figureFileBig=O+n5IugmpP9HKyRkkyUmBA==, tableContent=null), ArticleFig(id=1325495635268481382, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 15, caption=Physical topology diagram of the experimental environment, figureFileSmall=vcZPc6CjJoFI0d2wanBBUg==, figureFileBig=8cU0bZNsdv1KORQc564HiQ==, tableContent=null), ArticleFig(id=1325495635356561767, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图15, caption=实验环境的物理拓扑图, figureFileSmall=vcZPc6CjJoFI0d2wanBBUg==, figureFileBig=8cU0bZNsdv1KORQc564HiQ==, tableContent=null), ArticleFig(id=1325495635432059240, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 16, caption=Hardware performance of CXL JBOM, figureFileSmall=zQ2DLq2Hs+N5Gk62h+EpGQ==, figureFileBig=YcC074o/UY33i4LUcZK8cw==, tableContent=null), ArticleFig(id=1325495637109780842, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图16, caption=CXL JBOM 的硬件性能, figureFileSmall=zQ2DLq2Hs+N5Gk62h+EpGQ==, figureFileBig=YcC074o/UY33i4LUcZK8cw==, tableContent=null), ArticleFig(id=1325495637210444139, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 17, caption=memcpy performance of DRAM and CXL memory, figureFileSmall=Vy1pLVnwUuhCYonVtgIsPw==, figureFileBig=g9Xh31MpqMue8DKjpxHbWA==, tableContent=null), ArticleFig(id=1325495637315301740, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图17, caption=DRAM和CXL内存的memcpy性能, figureFileSmall=Vy1pLVnwUuhCYonVtgIsPw==, figureFileBig=g9Xh31MpqMue8DKjpxHbWA==, tableContent=null), ArticleFig(id=1325495637424353645, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 18, caption=Test result of applying memory, figureFileSmall=cy3H0rgOlRiPvOT2OA1P+w==, figureFileBig=noqbU6lqBuzBUJsS4yvifg==, tableContent=null), ArticleFig(id=1325495637512434030, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图18, caption=申请内存的测试结果, figureFileSmall=cy3H0rgOlRiPvOT2OA1P+w==, figureFileBig=noqbU6lqBuzBUJsS4yvifg==, tableContent=null), ArticleFig(id=1325495637596320111, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Figure 19, caption=Test result of CXL global cache performance, figureFileSmall=yMAhWRXgNDj4/3wER5FwqA==, figureFileBig=V6cIibY+kWY5xHP6s3QIsg==, tableContent=null), ArticleFig(id=1325495637709566320, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=图19, caption=CXL全局缓存性能测试结果, figureFileSmall=yMAhWRXgNDj4/3wER5FwqA==, figureFileBig=V6cIibY+kWY5xHP6s3QIsg==, tableContent=null), ArticleFig(id=1325495637797646705, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Table 1, caption=Advantages and Disadvantages of Typical Global Cache Technologies, figureFileSmall=null, figureFileBig=null, tableContent=
特点循环镜像双控横向扩展SCM全局缓存
性能高高低
高可用依次n坏n−1任意双控中的1个故障同时n坏n−1
扩展性4控制器可横向扩展可横向扩展
技术复杂度高中低
), ArticleFig(id=1325495637881532786, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=表1, caption=

典型全局缓存技术的优缺点

, figureFileSmall=null, figureFileBig=null, tableContent=
特点循环镜像双控横向扩展SCM全局缓存
性能高高低
高可用依次n坏n−1任意双控中的1个故障同时n坏n−1
扩展性4控制器可横向扩展可横向扩展
技术复杂度高中低
), ArticleFig(id=1325495638003167603, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Table 2, caption=Bandwidth of PCIe and RoCE, figureFileSmall=null, figureFileBig=null, tableContent=
高速互连技术带宽说明
PCIe 5.0 x8单向32 GBpsCXL 2.0
PCIe 5.0 x16单向64 GBpsCXL 2.0
PCIe 6.0 x16单向128 GBpsCXL 3.0
RoCE100 Gbps市场存储产品
RoCE200 Gbps在研存储产品
RoCE400 Gbps新的RoCE
RoCE800 Gbps下一代RoCE
), ArticleFig(id=1325495638137385332, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=表2, caption=

PCIe和RoCE的带宽

, figureFileSmall=null, figureFileBig=null, tableContent=
高速互连技术带宽说明
PCIe 5.0 x8单向32 GBpsCXL 2.0
PCIe 5.0 x16单向64 GBpsCXL 2.0
PCIe 6.0 x16单向128 GBpsCXL 3.0
RoCE100 Gbps市场存储产品
RoCE200 Gbps在研存储产品
RoCE400 Gbps新的RoCE
RoCE800 Gbps下一代RoCE
), ArticleFig(id=1325495638212882805, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=EN, label=Table 3, caption=Availability of Storage Clusters with Different Global Cache Designs, figureFileSmall=null, figureFileBig=null, tableContent=
编号集群/全局缓存设计控制器数量代表厂商可用性水平系统可用性/%
11框2控2多数厂商2坏199.999999
21框4控4多数厂商4坏199.999994
34控,分离式CXL缓存4本文4坏399.9…9小数点后16个9
48控,分离式CXL缓存8本文8坏799.9…9小数点后32个9
51框2控,横向可扩展32EMC框内2坏1,任何存储框不可故障99.99998
61框4控,8框集群32多数厂商框内4坏1,任何存储框不可故障99.999952
71框4控,8框集群,框间3副本32华为框内4坏1,存储框8坏199.99999999998992
), ArticleFig(id=1325495638300963190, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, language=CN, label=表3, caption=

不同全局缓存设计的存储集群可用性

, figureFileSmall=null, figureFileBig=null, tableContent=
编号集群/全局缓存设计控制器数量代表厂商可用性水平系统可用性/%
11框2控2多数厂商2坏199.999999
21框4控4多数厂商4坏199.999994
34控,分离式CXL缓存4本文4坏399.9…9小数点后16个9
48控,分离式CXL缓存8本文8坏799.9…9小数点后32个9
51框2控,横向可扩展32EMC框内2坏1,任何存储框不可故障99.99998
61框4控,8框集群32多数厂商框内4坏1,任何存储框不可故障99.999952
71框4控,8框集群,框间3副本32华为框内4坏1,存储框8坏199.99999999998992
)], 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/issn1000-1239.202550376, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202550376, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202550376, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202550376, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1772294400000, fullTextJson=null, articleText=null, reference=null)
收藏切换
基于CXL的存储阵列全局缓存技术
李强 1 , 刘涛 1 , 张雪庆 1, 2, * , 孟宪伟 3 , 孙明刚 1 , 李博乐 3 , 陈曦 1
计算机研究与发展 | 面向新质生产力的数据存储与处理专题 2026,63(3): 567-584
收起
收藏切换
计算机研究与发展 |面向新质生产力的数据存储与处理专题 2026 , 63 (3) : 567 -584
基于CXL的存储阵列全局缓存技术
全屏
[Author(id=1325495625223123229, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=li.qiangbj@ieisystem.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1325495625307009311, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625223123229, language=EN, stringName=Qiang Li, firstName=Qiang, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101, bio={"content":"

born in 1982. PhD, professor. His main research interests include computer architecture, storage system, and memory pooling

"}, bioImg=null, bioContent=

born in 1982. PhD, professor. His main research interests include computer architecture, storage system, and memory pooling

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495625390895392, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625223123229, language=CN, stringName=李强, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1浪潮电子信息产业股份有限公司 济南 250101, bio={"img":"Zo5PAPNV2Sv+zLmXuAiJTw==","content":"

1982年生。博士,正高级工程师。主要研究方向为计算机体系结构、存储系统、内存池化

"}, bioImg=Zo5PAPNV2Sv+zLmXuAiJTw==, bioContent=

1982年生。博士,正高级工程师。主要研究方向为计算机体系结构、存储系统、内存池化

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)])]), Author(id=1325495625546084642, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, orderNo=1, 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=1325495625646747940, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625546084642, language=EN, stringName=Tao Liu, firstName=Tao, middleName=null, lastName=Liu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101, bio={"content":"

born in 1983. Master, senior engineer. His main research interests include centralized storage architecture and hardware design

"}, bioImg=null, bioContent=

born in 1983. Master, senior engineer. His main research interests include centralized storage architecture and hardware design

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495625730634021, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625546084642, language=CN, stringName=刘涛, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1浪潮电子信息产业股份有限公司 济南 250101, bio={"img":"bRdcW4ZpWcmb+V3JzG8xMw==","content":"

1983年生。硕士,高级工程师。主要研究方向为集中式存储架构、硬件设计

"}, bioImg=bRdcW4ZpWcmb+V3JzG8xMw==, bioContent=

1983年生。硕士,高级工程师。主要研究方向为集中式存储架构、硬件设计

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)])]), Author(id=1325495625986486567, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=zhangxq@ieisystem.com, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1325495626078761258, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625986486567, language=EN, stringName=Xueqing Zhang, firstName=Xueqing, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101
2School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074, bio={"content":"

born in 1984. PhD candidate, assistant professor. His main research interests include storage system and system architecture

"}, bioImg=null, bioContent=

born in 1984. PhD candidate, assistant professor. His main research interests include storage system and system architecture

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495626422694187, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495625986486567, language=CN, stringName=张雪庆, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, 2, *, address=1浪潮电子信息产业股份有限公司 济南 250101
2华中科技大学计算机科学与技术学院 武汉 430074, bio={"img":"q/mOH6dy/SyvQfJq79akTA==","content":"

1984年生。博士研究生,助理研究员。主要研究方向为存储系统、系统架构

"}, bioImg=q/mOH6dy/SyvQfJq79akTA==, bioContent=

1984年生。博士研究生,助理研究员。主要研究方向为存储系统、系统架构

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)]), AuthorCompany(id=1325495625000825110, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=2, ext=[AuthorCompanyExt(id=1325495625013408023, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625000825110, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074), AuthorCompanyExt(id=1325495625021796632, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625000825110, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2华中科技大学计算机科学与技术学院 武汉 430074)])]), Author(id=1325495626540134701, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=1325495626812764463, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495626540134701, language=EN, stringName=Xianwei Meng, firstName=Xianwei, middleName=null, lastName=Meng, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095, bio={"content":"

born in 1986. Master. His main research interests include storage system design and memory pooling

"}, bioImg=null, bioContent=

born in 1986. Master. His main research interests include storage system design and memory pooling

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495626909233456, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495626540134701, language=CN, stringName=孟宪伟, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3浪潮(北京)电子信息产业有限公司 北京 100095, bio={"img":"ZVhoNEt/uPF/eD8JT51gYg==","content":"

1986年生。硕士。主要研究方向为存储系统设计、内存池化

"}, bioImg=ZVhoNEt/uPF/eD8JT51gYg==, bioContent=

1986年生。硕士。主要研究方向为存储系统设计、内存池化

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495625109877017, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=3, ext=[AuthorCompanyExt(id=1325495625122459930, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095), AuthorCompanyExt(id=1325495625135042843, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3浪潮(北京)电子信息产业有限公司 北京 100095)])]), Author(id=1325495627332858162, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=1325495627458687284, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495627332858162, language=EN, stringName=Minggang Sun, firstName=Minggang, middleName=null, lastName=Sun, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101, bio={"content":"

born in 1984. Bachelor, senior engineer. Member of CCF. His main research interests include storage software architecture and storage system performance

"}, bioImg=null, bioContent=

born in 1984. Bachelor, senior engineer. Member of CCF. His main research interests include storage software architecture and storage system performance

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495627706151221, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495627332858162, language=CN, stringName=孙明刚, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1浪潮电子信息产业股份有限公司 济南 250101, bio={"img":"4uNp/iUQuMLZ4KmcS6MLgw==","content":"

1984年生。学士,高级工程师。CCF会员。主要研究方向为存储软件架构、存储系统性能

"}, bioImg=4uNp/iUQuMLZ4KmcS6MLgw==, bioContent=

1984年生。学士,高级工程师。CCF会员。主要研究方向为存储软件架构、存储系统性能

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)])]), Author(id=1325495627815203127, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=1325495627915866425, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495627815203127, language=EN, stringName=Bole Li, firstName=Bole, middleName=null, lastName=Li, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095, bio={"content":"

born in 1984. Master. His main research interests include storage technology and hardware design

"}, bioImg=null, bioContent=

born in 1984. Master. His main research interests include storage technology and hardware design

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495628062667066, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495627815203127, language=CN, stringName=李博乐, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=3, address=3浪潮(北京)电子信息产业有限公司 北京 100095, bio={"img":"q1xKw1CBv+lr2AyQqSnCvg==","content":"

1984年生。硕士。主要研究方向为存储技术、硬件设计

"}, bioImg=q1xKw1CBv+lr2AyQqSnCvg==, bioContent=

1984年生。硕士。主要研究方向为存储技术、硬件设计

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495625109877017, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=3, ext=[AuthorCompanyExt(id=1325495625122459930, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095), AuthorCompanyExt(id=1325495625135042843, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495625109877017, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=3浪潮(北京)电子信息产业有限公司 北京 100095)])]), Author(id=1325495628133970236, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, 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=1325495628222050622, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495628133970236, language=EN, stringName=Xi Chen, firstName=Xi, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1IEIT SYSTEMS Co., Ltd., Jinan 250101, bio={"content":"

born in 1993. Master. Member of CCF. His main research interests include server system architecture and hardware design

"}, bioImg=null, bioContent=

born in 1993. Master. Member of CCF. His main research interests include server system architecture and hardware design

, aboutCorrespAuthor=null), CN=AuthorExt(id=1325495628477903167, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, authorId=1325495628133970236, language=CN, stringName=陈曦, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1浪潮电子信息产业股份有限公司 济南 250101, bio={"img":"lVlFZvsrLedDE2Ipq5l1rQ==","content":"

1993年生。硕士。CCF会员。主要研究方向为服务器系统架构、硬件设计

"}, bioImg=lVlFZvsrLedDE2Ipq5l1rQ==, bioContent=

1993年生。硕士。CCF会员。主要研究方向为服务器系统架构、硬件设计

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1325495624858218771, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, xref=1, ext=[AuthorCompanyExt(id=1325495624866607380, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1IEIT SYSTEMS Co., Ltd., Jinan 250101), AuthorCompanyExt(id=1325495624874995989, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1325495615110656219, companyId=1325495624858218771, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1浪潮电子信息产业股份有限公司 济南 250101)])])]
李强1 , 刘涛1, 张雪庆1, 2, * , 孟宪伟3, 孙明刚1, 李博乐3, 陈曦1
作者信息
  • 1浪潮电子信息产业股份有限公司 济南 250101
  • 2华中科技大学计算机科学与技术学院 武汉 430074
  • 3浪潮(北京)电子信息产业有限公司 北京 100095
通讯作者:
张雪庆()
作者简介:

1982年生。博士,正高级工程师。主要研究方向为计算机体系结构、存储系统、内存池化

1983年生。硕士,高级工程师。主要研究方向为集中式存储架构、硬件设计

1984年生。博士研究生,助理研究员。主要研究方向为存储系统、系统架构

1986年生。硕士。主要研究方向为存储系统设计、内存池化

1984年生。学士,高级工程师。CCF会员。主要研究方向为存储软件架构、存储系统性能

1984年生。硕士。主要研究方向为存储技术、硬件设计

1993年生。硕士。CCF会员。主要研究方向为服务器系统架构、硬件设计

Storage Arrays Global Cache Technology Based on CXL
Qiang Li1 , Tao Liu1, Xueqing Zhang1, 2, * , Xianwei Meng3, Minggang Sun1, Bole Li3, Xi Chen1
Affiliations
  • 1IEIT SYSTEMS Co., Ltd., Jinan 250101
  • 2School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074
  • 3Inspur (Beijing) Electronic Information Industry Co., Ltd., Beijing 100095
  • born in 1982. PhD, professor. His main research interests include computer architecture, storage system, and memory pooling

    born in 1983. Master, senior engineer. His main research interests include centralized storage architecture and hardware design

    born in 1984. PhD candidate, assistant professor. His main research interests include storage system and system architecture

    born in 1986. Master. His main research interests include storage system design and memory pooling

    born in 1984. Bachelor, senior engineer. Member of CCF. His main research interests include storage software architecture and storage system performance

    born in 1984. Master. His main research interests include storage technology and hardware design

    born in 1993. Master. Member of CCF. His main research interests include server system architecture and hardware design

出版时间: 2026-03-01 doi: 10.7544/issn1000-1239.202550376
文章导航
收藏切换

全局缓存是企业级存储阵列的关键技术,其核心功能是提高数据的写入速度:用户I/O写入存储控制器的内存缓存后立即返回,无需等待数据持久化至低速HDD或SSD设备。然而,企业级存储产品的全局缓存设计需综合考虑副本数量、系统高可用及故障场景等因素。受这些因素的相互制约,存储集群的控制器数量通常难以超过4个,且系统可用性无法支持4个控制器同时坏3个。此外,控制器间通过RDMA(remote direct memory access)传输缓存副本数据,也会影响存储性能。为此,基于最新的CXL(compute express link)技术,设计了内存分离式的全局缓存系统原型。内存缓存从控制器内部解耦至外部后,系统具有以下优势:控制器与内存缓存可独立横向扩展,可以支持多个缓存副本;控制器故障不会导致缓存数据丢失,n个控制器可以同时故障n−1个;CXL技术在提升系统扩展性与可用性的同时,能显著优化存储性能。实验结果表明,CXL全局缓存性能比RDMA全局缓存提升38%。

企业存储  /  存储阵列  /  CXL  /  全局缓存  /  分离式架构

Global cache is a core technology for enterprise-grade storage arrays, whose primary function is to improve data write performance. I/O operations return immediately once data is written into the memory cache of storage controllers, without waiting for the data to be persisted to slow HDDs or SSDs. However, the design of global cache in enterprise-grade storage products must comprehensively consider factors such as the number of replicas, system high availability, and failure scenarios. Constrained by the mutual interactions of these factors, the number of controllers in storage clusters typically cannot exceed 4, and system high availability fails to support the scenario where three out of four controllers fail concurrently. Furthermore, the transmission of cache replica data between controllers via RDMA(remote direct memory access) technology also impairs storage performance. Thus, a memory-disaggregated global cache system prototype is designed based on the latest CXL(compute express link) technology. With memory cache disaggregated from within controllers to external, the system offers three key advantages: first, controllers and memory cache can scale out independently, supporting multiple cache replicas; second, controller failures do not cause cache data loss, supporting concurrent failures of n−1 out of n controllers; third, CXL technology significantly optimizes storage performance while enhancing system scalability and high availability. Experimental results demonstrate that the CXL-based global cache achieves a 38% performance improvement compared with the RDMA-based global cache.

enterprise storage  /  storage array  /  CXL  /  global cache  /  disaggregate architecture
李强, 刘涛, 张雪庆, 孟宪伟, 孙明刚, 李博乐, 陈曦. 基于CXL的存储阵列全局缓存技术. 计算机研究与发展, 2026 , 63 (3) : 567 -584 . DOI: 10.7544/issn1000-1239.202550376
Qiang Li, Tao Liu, Xueqing Zhang, Xianwei Meng, Minggang Sun, Bole Li, Xi Chen. Storage Arrays Global Cache Technology Based on CXL[J]. Journal of Computer Research and Development, 2026 , 63 (3) : 567 -584 . DOI: 10.7544/issn1000-1239.202550376
以人工智能、大数据、云计算等战略性新兴产业为代表的新质生产力,对数据存储的访问性能、高可用性、可扩展性等方面提出了更加极致的需求。本文基于最新的CXL(compute express link)[1]技术,构建了新型的内存分离式全局缓存系统,在企业级存储阵列的先进数据分析处理技术领域进行了探索。
CXL是Intel公司在2019年推出的开放性互连协议,用于处理器、内存扩展和加速器的高速缓存一致性互连协议。CXL技术可以构建分离式的内存池。CPU,GPU等处理器可以通过同步的load/store指令高效访问远端内存。而且,CXL内存的时延很低,仅为250~500 ns,远低于RDMA的硬件时延2~3 μs。因此,CXL是目前性能最好的远程内存访问技术。
全局缓存[2-3]是企业存储的核心技术之一,最早出现在20世纪90年代的磁盘阵列产品上。其设计初衷是对HDD进行写加速,用户的I/O在写入DRAM缓存之后就会立即返回给用户,而无需等待长达10 ms的磁盘写入操作完成后再返回。该技术在当前的NVMe(non-volatile memory express)全闪阵列中也被广泛应用。
在高性能全闪存阵列中,通过RDMA技术在控制器间传输全局缓存的写I/O数据,会严重影响存储的性能。以浪潮存储系统为例,使用DRAM内存替代RDMA远程内存做缓存副本,同时保留相关软件功能,写I/O的IOPS从16.37万提升到了22.84万,提高了39.5%1。
高可用设计是全局缓存技术的另一个技术难点。当存储集群中的某个控制器断电或故障后,为了避免故障控制器中的DRAM数据丢失,全局缓存中的备份数据负责把用户的I/O数据写入到HDD或SSD中。为此,需要解决数据副本数量、副本如何在不同数量的控制器中分布、控制器故障和恢复等技术难题。当前最常见的实现方法是存储控制器间通过RDMA进行高速互连,2个数据副本在4个存储控制器中构成循环镜像结构。工程上的难点在于4个控制器的缓存镜像实现非常复杂,要考虑各种异常情况。因此,市场上没有超过4个控制器的全局缓存存储产品。
基于CXL的远程内存技术可以解决全局缓存的性能和高可用技术难题。从性能的角度来看,CXL比RDMA时延更低,能消除控制器间的数据流量,同时还能实现控制器外部的写缓存数据备份。从高可用角度来看,CXL把内存从控制器内部解耦到控制器外部,控制器故障和CXL缓存数据的故障相互隔离,两者互不影响。同时,存储软件无需考虑数据副本的分布问题,也无需在控制器故障后改变多控缓存系统的拓扑结构,控制器可以扩展至8个。
本文在浪潮信息存储产品AS18000上实现了系统原型,主要贡献有4个方面:
1) 在存储产品上引入最新的CXL硬件,提出了一种内存解耦的全局缓存架构,实现了原型系统,验证了该架构的技术可行性。
2)通过libnuma的函数numa_alloc_onnode申请CXL内存时延高达19.9 μs,本文开发了用户态内存管理模块,把申请内存时延降至1.31 μs。
3)CXL全局缓存技术显著提升了存储的性能,实验数据表明100%写性能提高了38%。
4)CXL全局缓存在现有的硬件上可以支持8控集群。支持n个控制器的同时坏n−1个,好于现有的RDMA循环镜像技术依次坏n−1个控制器。
为实现高可用性与高性能服务,企业级存储阵列通常采用多控制器设计。中低端存储采用双控制器设计,高端存储是4控。全局缓存具有以下技术价值:
1) 高可用。当存储系统中有任何1个控制器故障时,另外1个控制器可以接替其工作,从而保障用户读写数据的连续性和可用性。
2) 数据保护。用户数据写入时,双控会将数据同时写入2个控制器的缓存中,然后再写入到HDD或SSD中,这样即使1个控制器出现故障时,另外1台控制器中有数据备份,用户数据不会丢失。
3) 高性能。双控设计可以实现负载均衡,将数据访问请求均匀地分配到2个控制器上,避免单个控制器负载过高,从而提高系统的整体性能和响应速度。
下面介绍常见的全局缓存架构和技术。
最常见的全局缓存技术是双副本的循环镜像[4-5],即数据副本以循环的结构分布在各个控制器中,如图1所示。
该架构设计的好处:一是,正常I/O时仅多1次RDMA的网络传输,性能高。二是,当控制器数量变化时,比如1个控制器故障,集群控制器数量从4个变成3个时,副本的循环结构不变。
该架构的缺点如下:
1) 控制器数量超过4个时,故障处理会非常复杂,厂商一般会在4控的基础上进一步做更大规模集群,如8个4控互连形成32控集群,而不会做1个32控的大集群。
2) 当用户提出更高的可用性要求,比如缓存3副本,需要单独设计实现方式。比如华为公司把第3副本放在了另外1个4控集群中[5],但是跨框的网络通信会影响存储性能。
3) 2个副本无法支持控制器同时4坏3,只能支持依次4坏3。当1个控制器故障时,需要等系统完成缓存数据重建,形成3个控制器的循环镜像后,才能允许下一个控制器的故障。如果系统中有2个控制器同时故障,用户数据一定会丢失。
此外,基于DRAM的写缓存还需要独立的电池和掉电保护软件,当系统断电时,在独立电池的支持下可以把DRAM中的写缓存数据转移到SSD上,以避免数据丢失。
现在,浪潮信息[4]、华为[5]两家公司的高端存储产品都在使用该技术,国际厂商如EMC,HPE在2020年以前也在使用该技术。
由于循环镜像在多控制器场景下,软件设计和实现的难度太大,系统的扩展性受限。因此,2018年,EMC的PowerMax产品[6]改变了全局缓存的架构设计,变成了双控制器的横向扩展架构,如图2所示。
存储控制器内依然使用DRAM写缓存,控制器间依然使用RDMA传输I/O。与循环镜像不同的是,写缓存只在每个存储框中的2个控制器间做镜像,写缓存镜像数据不会写入第3个控制器中。其结果是把全局缓存的高可用设计简化为2个控制器。
该架构的优点是存储的集群规模与全局缓存的复杂度无关,系统可以横向扩展。PowerMax最多可以支持16个存储框和32个存储控制器。
该架构设计的缺点是只能支持1个框中的1个控制器故障,如果1个存储框中的2个控制器同时故障,用户的写入数据会丢失。EMC的全局缓存设计是牺牲了系统的可用性来提高了存储集群的扩展性,用更大规模的存储集群来对外提供高性能。
2019年成立的VastData[7]公司在全局缓存技术上进行了创新。VastData采用了非易失内存(storage class memory,SCM)和NVMe-oF(NVMe over fabric)技术替代了DRAM和RDMA。写缓存被保存在存储控制器外部的SCM上,任意一个控制器可以通过NVMe-oF协议来访问任意一块SCM上的写缓存数据,如图3所示。
基于SCM的全局缓存的优点是:
1) 缓存数据被解耦到了控制器外部,控制器故障不会导致写数据的丢失,硬件上可以做到控制器n坏n−1的高可用。
2) SCM是非易失介质,存储系统内无需增加单独的供电电池和无需掉电保护软件。
3) 避免了复杂的循环镜像技术。
该技术有2个缺点。一是性能差。SCM的性能不如DRAM,NVM-oF的性能不如RDMA。二是现在市场上没有好的SCM产品。Intel optane persistent memory在2024年12月31日后不再接受订单,2025年12月31日后停止供货,虽然市场上还有三星的Z-NAND和铠侠的XL-Flash可以作为optane的替代品,但其性能均不如optane。
HPE的Alletra MP B10000[8](2025年以前,产品名称为greenLake for block storage)则更为激进,直接去掉了SCM和其他形式的全局缓存,所有I/O直接写入SSD,该方法的缺点是写性能更低。
表1列出了以上3种全局缓存技术的优缺点。
本节介绍CXL的相关产品和技术现状。虽然2019年就已经发布了CXL 1.0规范,但是,从2022年才开始出现CXL的相关硬件产品,包括芯片、CPU、内存设备、交换机和CXL SSD。
1) 芯片。澜起科技(Montage Technology)公司在2022年发布了世界首款CXL内存扩展控制器芯片MXC[9],该芯片支持JEDEC DDR4和JEDEC DDR5标准,符合CXL 2.0规范,支持PCIe 5.0的速率。
2) CPU。2023年初,Inter公司发布了支持CXL 1.1的SPR(sapphire rapids) CPU,随后又发布了支持CXL 1.1的EMR(emerald rapids)CPU。2024年第3季度,Inter公司发布了2款支持CXL 2.0协议的CPU,granite rapids和sierra forest(已更名为“至强6”)。AMD在2023年发布了支持CXL 1.1协议的genoa CPU,2024年第2季度发布了支持CXL 2.0协议的第5代EPYC处理器“Turin”。
3) 内存设备。包括AIC(add in card)扩展卡、E3.S形态的内存模组和内存簇(just a bunch of memory, JBOM):2022年,澜起科技公司在开发计算项目全球峰会(Open Compute Project Global Summit, OCP Global Summit)上发布了CXL AIC卡的评测板[10]。2022年,三星公司发布了容量为512 GB、形态为E3.S的CXL内存模组[11]。2024年,三星公司发布了名为CMM-B的CXL内存盒模组[12],浪潮信息公司自研了总容量为16 TB的JBOM原型产品Amber[13]。
4) 交换机。2023年8月,XConn公司发布了首个支持CXL 2.0的交换机[14]。
5) CXL SSD盘。三星公司在2023年发布了CXL SSD[15],该SSD提供了2种接口CXL.mem和CXL.io,用于访问盘内部的DRAM内存和NAND,并具备3种模式:在非易失模式下,应用通过load/store语义访问盘内的DRAM,当服务器断电时,由外部电池供电,服务器负责转移DRAM中的数据到盘内的NAND中;在NAND模式下,服务器使用CXL.io和NVMe协议直接访问盘内的NAND;在缓存模式下,服务器通过load/store语义访问DRAM中的缓存数据,NAND做容量层,但是需要服务器自己做缓存的数据管理。该产品的成果发表在HotStorage 2022[16]。
2023年的闪存峰会(the Future of Memory and Storage,FMS)上,三星、MemVerge和H3等公司联合展示了2 TB的池化CXL内存系统解决方案Pooled CXL Memory System[17]。该方案最多可以支持8台服务器,允许主机按需使用内存资源。同时,还配备了用于可视化、池化、分层、动态内存分配的软件。
在学术领域,CXL相关的研究工作有:内存池、硬件测试、CXL SSD、跨机柜的CXL内存池、分层软件、共享软件、CXL应用等。
在硬件方面,文献[18-22]在云环境下使用CXL内存池以提高内存的利用率,降低数据中心的成本。文献[23-25]是系统测试工作,Liu等人[23]在5个CPU平台下测试了4个CXL设备的265个工作负载,多数负载可以容忍CXL内存的延迟,性能下降不明显,这和本文的测试结果一致,CXL全局缓存和DRAM内存环回性能相近。文献[16, 26-27]开展的是CXL SSD相关的研究工作。Wang等人[28]通过CXL和RDMA相结合的方式来实现跨机柜的大规模内存池。Wang等人[29]则是通过让CXL协议运行在以太网上来实现跨机柜的内存池。两者解决的都是CXL设备的传输距离约为1 m且只能在1个机柜内应用的问题。
在软件方面,文献[30-32]研究的是Linux系统中DRAM内存和CXL内存之间的高效分层,让更多的数据访问落在低时延的DRAM内存中。Zhong等人[33]在CXL 2.0上做了内存共享软件,采用了绕开CPU缓存的方法来解决2.0协议不支持缓存一致性的问题。Wang等人[34]研究哈希表的优化,通过存储更多元数据减少CXL内存访问。文献[35-36]研究的是把CXL内存应用到数据库。文献[37-41]应用大容量的CXL内存来解决AI训练过程中的性能问题。
在存储阵列上使用CXL技术,要解决3个问题:一是,要让存储系统支持CXL;二是,解决内存共享问题;三是,保证带宽够用,且不会成为系统瓶颈。针对第1点,我们开发了Retimer卡、系统固件和BIOS。本节分析CXL内存共享和带宽。
存储阵列是一个多控制器构成的小型集群系统,用户的I/O可以发送给任意一个控制器,然后控制器再访问全局缓存中的数据。多个控制器的CPU有可能在同一时间访问同一个CXL内存的物理地址,多个CPU的缓存有可能会保存同一个CXL数据的缓存,这产生了多个CPU缓存间以及CPU缓存与CXL内存间的缓存数据一致性问题。CXL 3.0协议[1]在MESI协议基础上设计了反向失效嗅探(back-invalidate snoop,BISnp)的机制来解决该问题:首先,CXL内存芯片会使用部分内存空间记录各个CPU缓存的数据缓存情况。当芯片嗅探到CPU发来的写操作时,CXL芯片会主动让所有缓存该数据CPU缓存都失效。然后,再允许CPU完成写操作,以此保证多个控制器的缓存和CXL内存的缓存数据一致性。
但是,现在还没有支持CXL 3.0的CPU和产品,CXL 2.0不支持缓存一致性。常见的解决方法有2种:一是绕开CPU缓存[42],如利用Intel CPU的MTRR(memory type range registers)[43],所有的内存数据都直接写入CXL内存。二是在向CXL写入数据前,用带外的方式,如UDP,协调多个控制器间的读、写权限,当有1个CPU写CXL内存时,其他CPU停止访问该数据,以此保证缓存数据的一致性,如Memverge的CXL内存共享软件Gismo[44]。
与上述2个方法不同,本文则是在CXL上层的存储软件利用数据分片来解决CXL内存共享问题。即让每个I/O进入到每个控制器前进行数据分片,保证每一个I/O只会被发送到某一个固定的控制器上,然后每个控制器再访问与自己相连的CXL内存。这样用户在使用存储的过程中,不会发生同一个I/O会经由不同控制器读、写同一个CXL物理内存的情况,也就消除了缓存不一致的现象。
具体做法是修改存储的服务器多路径软件,引入数据分片算法,在向存储发送I/O前,多路径软件根据I/O的LBA地址计算出目的控制器,然后发送I/O。这样保证了在任何时刻,同一个地址的I/O都会发往同一个控制器,每个控制器只会访问自己的缓存,避免了不同的控制器访问同一缓存数据的情况。
与DRAM相比,CXL内存没有带宽优势[23-25],即便是通过CXL交换机聚合了多条内存后,PCIe x8的CXL内存的单向带宽也仅为32 GBps,而DDR4 3200内存的单向带宽为25.6 GBps,DDR5 4800内存的单向带宽为38.4 GBps,因此,CXL内存带宽相当于1条内存的带宽。但是,即便这样,CXL内存的带宽也可以满足企业级盘阵存储的需求。我们从与RDMA带宽对比和系统性能瓶颈2方面来分析。
表2给出了PCIe 5.0和PCIe 6.0的CXL内存带宽以及RoCE(remote direct memory access over converged Ethernet)的带宽。可以看出,即便是本文使用的PCIe 5.0 x8的CXL内存带宽也达到了32 GBps,好于在研存储产品使用的RoCE 200 Gbps。
下面分析现有的企业盘阵产品的实际最大性能。存储性能委员会(Storage Performance Council,SPC)的2025年2月SPC-1基准评测榜[45]排名第1的存储产品,在39%随机读和61%写的混合负载下,32控制器总IOPS为3000.2765万,平均每个控制器是93.76万IOPS,32控制器的峰值带宽是350 GBps,平均每个控制器是10.9 GBps带宽。而PCIe 5.0 x8 CXL内存的单向物理带宽为32 GBps,实测带宽为21.6 GBps,完全可以满足盘阵的性能需求。
从上面的分析可知,虽然CXL内存的带宽不如DRAM,但其PCIe 5.0 x8的21.6 GBps的带宽仍然能满足存储的需求,并不会成为系统性能的瓶颈。另一方面,与现在广泛采用的RoCE技术相比,不管是现在还是将来,CXL都有带宽优势。
图4展示了基于CXL技术的全局缓存的设计。从系统拓扑图来看,CXL全局缓存更接近于VastData的SCM全局缓存设计,同样是计算-缓存的分离式架构,控制器可以独立横向扩展,所有控制器都能访问到全局缓存中的数据。不同的是用分离式的内存替代SCM以及控制器使用更高效的load/store语义和CXL协议访问内存,盘阵的性能会更高,甚至好于目前性能最好的RDMA全局缓存。
CXL内存把原位于控制器内部的缓存数据解耦到外部,具备4个优点:
1) 扩展性更强。计算资源不足时增加控制器数量;要提高系统可用性,可以增加JBOM和缓存副本数量,软件复杂度不变。CXL 2.0交换机最多支持8个控制器,存储集群最高支持8控。CXL 3.0协议支持交换机级联,对于存储来说集群规模几乎可以无限扩展。
2) 性能更高。在硬件方面,CXL的时延仅为百纳秒,好于RDMA的2~3 μs。在软件方面,CXL内存复制使用的是标准的函数memcpy,也可以采用汇编指令、Intel DSA(data streaming accelerator)硬件加速库进行性能优化。而RDMA则需要运行驱动程序,CXL的内存访问性能优势明显[22,28,46-48]。在构建分离式内存池方面,CXL在性能、池化、缓存一致性上优于RDMA[18,20]。在实际应用方面,阿里PolarDB使用CXL内存池后,吞吐量比RDMA内存池高2.1倍[49]。因此,基于CXL在企业存储阵列上构建全局缓存,性能会有明显的提升。
3) 可用性更高。传统存储使用控制器内的DRAM保存写缓存副本,如果是m副本的系统,最多支持m−1个控制器故障,否则就会有丢失写缓存数据的风险。控制器和内存解耦后,不管控制器故障多少个,外部JBOM中的数据都不会丢失。因此,n个控制器可以同时故障n−1个,写缓存数据不丢失,存储系统的服务不中断。另外,JBOM本身是纯硬件,没有存储软件,其故障率仅受硬件影响,系统故障率远低于控制器的故障率。
4) 故障处理过程比循环镜像更简单。无需考虑复杂的故障组合,故障处理软件的复杂度与控制器数量无关,解决了RDMA循环镜像的软件扩展难的问题。
下面我们从硬件、CXL内存管理、写缓存设计、持续镜像4个方面说明系统设计。
本文在澜起科技公司的CXL芯片[9]基础上,自研了CXL JBOM原型产品Amber[13]和OCP(open compute project)形态的Retimer卡来实现存储的CXL全局缓存硬件系统。
CXL JBOM是19寸2U标准机箱,对外提供16个x16端口,每个端口对应4条容量为256 GB的DDR5 4800 MTps内存,总共支持64条内存,最大容量为16 TB;单端口支持1 TB内存扩展;支持2个CRPS(common redundant power supply)电源,1+1冗余;外部线缆采用CDFP DAC(c form-factor pluggable direct attach cable)线缆互连。图5是实物图。
每台JBOM内部由2块内存板构成,单块内存板具备16个PCIe 5.0 x8 的澜起MXC芯片,每颗MXC芯片配1个SPI(serial peripheral interface)Flash用于存放芯片固件,支持2条内存实现内存拓展,单板总计支持32条内存。每2颗MXC的x8 CXL链路组成1x16的CXL链路,并通过CDFP连接器对外互连。CDFP连接器同时提供同源时钟以及PERST信号,用于MXC芯片时钟输入及复位。同时板卡提供非同源时钟方案,提供更可靠的输入时钟源冗余特性。单板同时提供debug连接器用于调试。图6为其硬件拓扑图。
服务器上一般会使用Retimer卡把PCIe接口转成CDFP接口,然后通过CDFP DAC线缆连接到CXL JBOM或者交换机上。同时,Retimer卡还能延长PCIe高速信号的传输距离。它通过重新定时(Retiming)信号来增强信号质量,避免传输过程中信号衰减导致的错误。经过Retimer卡后,控制器外部的线缆的可用长度从0.9 m延长到1.1 m。但是,因为企业级存储采用的是定制化硬件,内部空间有限,无法使用标准服务器Retimer卡来连接JBOM和存储的控制器。因此,我们自研了OCP形态的Retimer卡,图7是实物图。
高性能盘阵时延通常为数百微秒,低负载场景下可低至100 μs。而通过函数numa_alloc_onnode向内核申请4 KB CXL内存的时延达19.9 μs,难以满足高性能存储的需要。为此,本文设计了用户态内存管理模块,以减少内存分配的时间开销。
如图8所示,为满足存储的高性能需求,本文采用用户态方式统一管理CXL与DRAM内存。CXL与DRAM内存管理模块相互独立,内部软件设计则完全一致,差异仅在于CXL模块的函数、变量名及API接口均带有“cxl_”前缀。
在系统初始化阶段,2个模块通过函数numa_alloc_onnode分别独立地向内核申请内存资源。随后,系统在用户态以4 KB为单位管理内存空间,并通过内存池机制向各软件模块提供预留的内存资源。
为了便于管理和扩展,内存管理模块设计成了2级页表,包括pd(page descriptor)、1级页表和2级页表,最大可管理2 TB的内存空间。
1) pd是指向4 KB内存页的指针,大小为8 B。
2) 1级页表由若干个4 KB的1级页表项构成,每个页表项保存512个指向2级页表项的pd。1级页表固定占用1 MB的空间,有256个1级页表项,可以保存128K个pd。
3) 2级页表由若干2级页表项构成,每个2级表项大小为32 KB,2级页表项内部由4096个pd构成,每个pd指向1个大小为4 KB的物理内存页。
4) 物理内存页每个内存页大小为4 KB,可以是CXL内存或者是DRAM内存。设计为4 KB的原因是企业硬盘的块大小一般是以4 KB为单位。每个物理内存页都在2级页表中对应1个pd以及1个页号。用户可以在2级页表数组中根据页号找到对应内存页的pd。
pd、1级页表、2级页表、物理内存页的关系如图9所示。
2级页表数据结构设计一共可以管理512M个pd和4 KB内存页,1个DRAM或CXL内存管理模块总共能管理2 TB内存,2级页表最多占用额外的4 GB内存,1级页表占用固定的1MB空间。
用户通过页面号访问4 KB内存页,假设用户给的页面号为p,1级表项的变量名为level1_memorytable。 当系统初始化完毕时,用户拿到的2级页表项是一个连续的空间,因此获得p对应的4 KB页面指针的伪代码为:
为了管理和使用DRAM或CXL的内存页空间,我们把所有4 KB的内存页划分成了5个区域:pd区域、general pages区域、空闲页(free pages,FPS)区域、IOB区域(32 KB的大IO块)和anchor区域。其中,general pages区域是业务模块能用到的内存空间,其他都是必要的管理general pages所需的空间。
1) pd区域。对应1级和2级页表,总管所有的内存页。软件以1级页表为入口,可以访问到下述所有的内存区域。
2) general pages区域。业务模块使用的4 KB内存空间。当业务模块向DRAM或CXL内存管理模块申请时,获得的是该区域的内存地址,存储软件运行过程中使用的也是该内存区域。
3) 空闲页区域。管理“空闲”内存页,用于用户态内存的分配和回收。
4) IOB区域(IOBlock)。向上层模块提供32 KB的大块内存。每个IOB由8个4 KB内存页构成,每个IOB都有1个iob number。软件模块通过iob number拿到IOB,最后从pte中拿到需要的内存页。
5) anchor区域。保存的是内存管理模块的基础信息,包括可用页面数量、空闲页面数量、IOB数量,以及存储控制器id、集群信息等。anchor占用固定的32 KB空间。
存储I/O处理是一个高并发过程,存在很多竞争条件。每个存储软件模块在运行阶段以内存池为单位,预留内存可以有效降低竞争和死锁的概率。
本文采用了树形结构的多级内存池来管理内存资源,树的每个节点是一个内存池。内存池的关键数据结构有:存储资源对象的链表、维护资源的计数等。
每个节点池都有各自的配额,配额即从内存管理模块中预留的内存空间。根节点是该资源类型的总池,每个模块可以从根节点建立多级子池,子池的配额是上级池或总池的一部分。
图10展示了3级内存池的结构。通常情况下,资源申请都在内存池的配额内,纯用户态运行,效率非常高。但是,当某个池上的资源申请超过其配额时,该池就会向父节点的内存池借用内存页,用完了再还回去。如果一直到根节点都没有借到,就要等待其他人释放了才能申请成功,这样就允许超配额使用资源,避免配额成为系统瓶颈。
由于企业级存储有高可用的需求,写缓存至少要存2份,以保证当1个数据缓存出现硬件故障后,数据不丢失。采用CXL内存后,有2种缓存双副本方案:方案1是2个CXL JBOM各存1个副本;方案2是DRAM内存1个副本和CXL1副本。
在现阶段,方案2具有明显的优势。理由如下:
1) DRAM的性能好于CXL,时延和带宽都优于CXL。但是,这不是主要因素。因为,硬件时延只相差150~350 ns,memcpy的总时间相差不到2 μs,这对于总时延接近1 ms的存储系统来说,影响不大。前文分析过带宽,CXL带宽相比DRAM带宽虽然不高,但不是存储系统的性能瓶颈。
2) CXL不支持DMA(direct memory access)。企业存储在访问前、后端网卡,或者访问控制器内的SSD盘时,使用的都是DMA。但是,现在的DMA并不支持CXL。因此,如果要维持之前系统的功能,使用网卡的DMA,写数据时要先由网卡写入到DRAM,然后用memcpy复制到CXL内存。读数据时,先把CXL内存中的缓存数据复制到DRAM中,然后才能返回给网卡。在落盘时也是一样的过程。很明显,这个过程中,各多了1次内存拷贝。
为了提高性能和减少不必要的数据拷贝,我们采用了方案1 DRAM+CXL的方式。用户的写I/O先经过网卡的DMA写入到DRAM,然后软件调用函数memcpy复制1份到CXL内存。后续缓存数据落盘时,把DRAM的缓存数据写入到SSD中,在正常I/O过程中,该方式不会带来额外的内存复制开销。CXL内存起到备份作用,当控制器故障时,CXL内存中的缓存数据保证I/O可以继续进行,用户数据不丢失。
与使用2个CXL内存的方案1相比,方案2的劣势是要同时管理DRAM和CXL缓存数据,软件实现相对复杂。
图11对比了RDMA全局缓存和CXL全局缓存的I/O处理过程。图12是RDMA缓存镜像的I/O流程图。本节通过对比,说明CXL全局缓存的优势。下面简述CXL全局缓存过程。
① 主机经过数据分片的Hash计算,知道该I/O应发往控制器1。
② 写缓存模块把I/O保存在DRAM中,然后通过函数memcpy把数据复制到CXL内存中,同时调用系统,把这段数据刷写到CXL内存中,不要仅仅保存在自己的CPU 缓存里。此时1份数据保存在DRAM内存中,另一份数据保存在CXL内存中。
③ 控制器1响应主机:数据写入成功。
④ 开始把写缓存数据异步刷写到硬盘。
⑤ 异步刷写成功,内存中的数据被写入到了硬盘中。
⑥ 脏数据转非脏:DRAM的数据已经写到了硬盘中,CXL中的副本数据无需再次下刷,但是要转换数据状态为非脏。
从以上流程可以看出,控制器1没有和其他控制器有任何的交互,如果把控制器增加到4个甚至8个,缓存的拓扑结构也不会变化,因此系统具备很好的横向扩展能力。图11右图的②中也只是执行了1次函数memcpy,系统开销小。与图12左侧的RDMA缓存镜像相比,控制流程大幅度简化,因此性能有较大幅度的提升。
需要进一步说明的是,图12左侧②和③的数据传输过程包含3次RDMA交互。之所以如此设计,是因为RDMA作为一种异步的、绕过对端CPU的高性能网络通信技术,与存储缓存镜像所需的同步机制存在差异,缓存镜像要求同步处理,且需要对端CPU参与本地数据处理。因此,需进行以下开发工作。
1) 异步改同步。当1份缓存数据成功写入到对端控制器的内存中后,需要告知发送端写入成功。
2) 缓存写入成功后,要唤醒对端的CPU进行相关逻辑的处理。实现上,我们采用的是轮训机制。
上述2个因素致使向对端写入缓存数据需经历 3 次RDMA传输,而且轮询操作会消耗多余的CPU资源。而CXL作为同步写入技术,只需CPU的store指令返回,就表明数据已成功写入CXL内存中,这极大简化了RDMA的缓存镜像处理流程。因此,CXL全局缓存的性能提升不仅仅是由硬件的2~3 μs降低到0.35 μs。
文献[46]阐述了面向4控集群,基于RDMA的持续镜像技术原理,实际的多控产品的情况要更复杂。
图13展示了4控存储系统发生节点依次故障时的拓扑视图变化。每次控制器故障,存储都要经历节点配对、传输镜像数据2个主要过程。配对过程是要为每个数据块找到新的配对节点,一般5 s以内完成,在此过程中I/O流会中断。配对完成后,原来的副本节点对外提供I/O服务。数据传输过程会把故障节点的几十到上百GB的缓存副本传输到新的镜像节点上。该过程耗时数十秒,并消耗很多CPU资源,新配对节点在完成数据传输后才能提供I/O服务。
此外,循环镜像还要处理其他相关复杂问题:及时下刷缓存脏数据;严格保证缓存副本数据的一致性;当节点恢复时,新加入的节点会多1个副本,要处理短暂的“3副本”状态;控制器重启会在短时间内引发2次重新结对,数据来回迁移2次,我们称之为“震荡”,软件上要设计防“震荡”策略。
如果存储集群规模变大,还需要解决跨框的“多控循环镜像”的难题。常见的做法是把8个控制器连在了一起,没有实现真正的跨框32控循环镜像。此时,如果1个框故障,缓存数据就会丢失。有的厂商在另外1个控制框中增加了第3副本[5],容忍了1个控制框的故障。缺点:1)实现复杂;2)多了1次框间数据传输,降低了存储性能。
CXL全局缓存可以大大简化多控缓存镜像的实现逻辑,如图14所示。当有1个或多个控制器同时故障时,集群的主控节点依据负载均衡的原则,让其他控制器直接接管故障控制器的CXL缓存数据即可实现,难度大大降低。同时,在这个过程中无需进行数据迁移或复制。
本文在理论上计算循环镜像、CXL全局缓存、32控集群等不同全局缓存架构的可用性,以说明CXL全局缓存在高可用方面的优势。表3汇总了全局缓存技术的存储集群可用性。
根据控制器或控制框的数量,以及同时允许故障的控制器数量来计算存储的可用性。n个控制器可以视作n次独立重复试验,每个节点的试验只有“故障”和“正常”两种结果,而且每个控制器的故障相互独立,而且故障概率均为p。因此,m个控制器同时故障的概率符合二项分布,可以用式(1)估算n个节点最多同时有m个故障的概率。多数存储产品都用缓存双副本,因此m=1。1框2控和1框4控存储的可用性均按照式(1)直接计算。
$ A=\sum _{i=0}^{m}{{\mathrm{C}}}_{n}^{i}{\left(p\right)}^{n-i}{(1-p)}^{i} \;。$
CXL全局缓存硬件虽然是由控制器和CXL设备构成,但是我们认为CXL设备是纯硬件,可用性极高,只需要根据控制器的可用性估算系统可用性。CXL全局缓存的可用性特点是n个控制器允许n−1个同时故障,其可用性采用式(2)计算。
$ A=1-{(1-p)}^{n} \;。$
表3第5行和第6行的32控存储中,单个控制框的可用性按照式(1)计算。由于框间没有写缓存数据副本,任何一个框的故障都会导致写I/O的数据丢失。因此,该设计下任何存储框都不能故障,其集群可用性计算为式(3),其中q为单个框的可用性。
$ A={q}^{n} \;。$
8框32控3副本存储(第7行)的实际效果是支持框内控制器4坏1,集群最多支持1个框故障。因此,可用性计算方法是先按照式(1)计算单个框的可用性,然后再次使用式(1)计算8框的可用性。
高端存储的可用性一般为5~6个9[4-8],单控制器可用性一般为4个9,即p = 99.99%。在此基础上,表3计算了不同控制器数量、不同全局缓存设计的集群可用性。
从表3可知,如果单节点的可用性均为4个9,CXL全局缓存的可用性是最高的,4控可达16个9,8控集群的理论可用性达32个9。CXL全局缓存支持n个控制器同时坏n−1,在所有的全局缓存设计中,其可用性水平也是最高的。
需要说明,可用性和可靠性是存储产品系统设计的一个方面,但也要考虑性能、容量、可扩展性和实现难度。更大规模的16控、32控集群的优点是:多控制器并发处理I/O会为用户提供更高的性能,更多的控制器能管理更多数量的硬盘,为用户提供更多的卷数量和更大的卷空间。因此,EMC在其高端产品中使用了1框2控横向可扩展的架构,没有采用可用性更高的框间3副本设计。框间3副本的问题:一是使用的DRAM内存数量多、成本高;二是相同规模的集群性能差,多复制1份数据会带来额外的数据传输,同时框间的传输速率更低。本文设计的CXL全局缓存能保证高可用的同时提升性能,有很强的技术先进性和应用价值。
本文基于浪潮信息存储产品实现了用户态内存管理模块与CXL全局缓存。实验环境部署了2台服务器、1台浪潮信息存储设备、2台JBOD及1台JBOM,硬件拓扑结构如图15所示。
测试服务器是2台服务器:一台CPU为Intel® Xeon® Platinum 8358,主频为2.6 GHz,内存为312 GB;另一台CPU为Intel® Xeon® Gold 6330,主频为2.0 GHz,内存为512 GB。2台服务器均使用16 GBps的FC卡与存储直连。
存储使用的是浪潮信息双控存储产品AS18000,每个控制器具有2路CPU和256 GB内存,CPU型号为INTEL® XEON® GOLD 6454S,主频为2.2 GHz,支持CXL 1.1,可以访问CXL 2.0的内存设备。
后端的存储扩展柜是2台共享JBOD,一共有40块1.92 TB的SAS SSD盘,型号为三星PM1643。
JBOM使用的是自研的Amber产品,使用的是SK海力士DDR5内存,每条内存的容量为128 GB,主频为5600 MHz。每个控制器通过1条CDFPDAC线缆连接到JBOM,可以访问到2条DDR5内存,共256 GB。
首先,我们使用Intel MLC(memory latency checker)工具[50]测试JBOM的硬件基础性能,测试结果如图16所示。JBOM的时延约是DRAM近端numa节点时延的3.36倍,远端CXL内存时延比近端时延多了158 ns。
存储缓存软件在访问CXL内存时,主要通过函数memcpy把4 KB缓存数据从DRAM复制到CXL中。 因此,本文开展了相关验证测试工作,将测试分为CPU缓存命中和缓存不命中2组,测试结果如图16所示。之所以设置2组测试,是因为在顺序写入或用户数据存在热点的情况下,CPU缓存会出现大量命中现象,这会导致函数memcpy的性能发生显著变化。图17是DRAM内存和CXL内存的memcpy函数测试结果。
图17(a)的测试结果表明,当CPU缓存命中时,DRAM内存和CXL内存的性能基本没有差别。图17 (b)展示了不同数据块大小、缓存不命中时的2种内存的memcpy性能。缓存不命中的测试方法为:在测试 CXL 内存前,先在1块超过CPU缓存总容量的大块内存上运行函数memcpy或memset,随后再执行DRAM或CXL内存的memcpy操作。由于此时CPU缓存已被大块内存数据完全覆盖,因此测得的时延值必然是缓存不命中的测试结果。此外,测试过程中需避免重复访问同一内存地址,若memcpy操作针对同一内存地址执行,时延会大幅降至0.3 μs。图17(b)的测试结果显示,通过调用函数memcpy后,DRAM与CXL的性能差距显著缩小。尽管CXL的硬件时延是DRAM的3.34倍,但在使用memcpy复制4 KB数据块时,CXL时延仅为DRAM的1.42倍。因此,在复杂软件系统中,若对CXL内存数据进行有限次的复制操作,并不会对系统性能造成影响。
图18展示了使用malloc、numa及本文4.2节开发的用户态内存管理模块进行内存申请的性能测试对比结果。测试数据表明:若通过 numa_alloc_onnode接口申请CXL内存,时延高达19.9074 μs,是malloc内存时延的4.73倍。该数值在总时延为100 μs的高端存储系统中占比过高,会显著影响存储性能。而通过用户态内存管理模块申请内存,时延可降至1.31 μs,优化效果显著。
我们采用金融、通信行业的存储集中采购测试方案,对比RDMA全局缓存、CXL全局缓存的性能,重点关注8 KB小I/O随机写场景下的IOPS性能。系统配置为:48个卷,每个卷500 GB,采用分布式RAID5,条带宽度为9,SSD盘数量为40。测试的软件采用vdbench 5.04.07,随机写,预热时间和测试时间都是5 min。性能评价方法为:调整vdbench的线程数,在保证时延不超过1 ms的条件下,存储IOPS的最大值即为测试结果。图19是全局缓存的性能测试结果。
图19(a)比较了无缓存、RDMA全局缓存、CXL全局缓存、内存环回的性能。无缓存代表了1.3节无缓存架构的性能值。RDMA缓存代表了1.1节循环镜像和1.2节双控横向扩展架构的性能值。内存环回指的是在本地DRAM中做数据副本,同时保留相关的软件逻辑,环回代表了双副本全局缓存的性能极限。测试结果表明RDMA全局缓存比无缓存的性能提升了35.29%,如果取消内存缓存,那么存储的性能会下降。CXL全局缓存的性能比RDMA全局缓存的性能提升了38%,同时仅比内存环回性能低0.24%,接近双副本系统的性能上限。
CXL全局缓存和内存环回性能相近的原因:一是CXL的软硬件时延和DRAM相近;二是在复杂的存储软件栈中,多几次CXL和DRAM的内存拷贝,对性能影响极小。
图19(b)展示了不同块大小的性能测试结果,性能提升幅度在30.7%~38%之间。图中64 KB带宽比32 KB下降的原因是线程数减少了一半。如果64 KB测试用例的线程数和32 KB的线程数一样,那么时延会显著超过1 ms,这不符合企业客户对全闪存储的测试要求。
虽然本文的研究工作证实了CXL全局缓存的技术优势和性能表现。但是,大规模产品化仍面临以下挑战。
1) 集成到存储内部。自研的CXL JBOM和交换机[13]均是独立的2U设备。而4控企业存储的高度为4U,与之相比CXL硬件占用的空间太大。因此,存储产品需要把JBOM和交换机都小型化并集成到内部。对于双控存储来说,因为写缓存容量不需要太大,只需要少量内存和2个CXL芯片,而且CXL芯片现在已支持双端口,无需额外配置交换机。因此,双控存储内部集成CXL是可行的。但是,4控、8控集群必须要使用交换机才能实现控制器4坏3和8坏7,需要在现有的存储上增加2U的高度,系统总高度将达到6U。
2) 热插拔。这是产品化过程中最具挑战性的环节。由于CXL设备位于存储外部,所以不管是人为拔除线缆,还是内存或CXL芯片故障,若没有热插拔功能,就会导致存储系统的崩溃。实现热插拔需要服务器、OS、存储软件协同解决:服务器硬件要能感知到CXL设备故障,然后生成中断给OS,OS完成内核的故障处理后通知存储软件,最后由存储软件完成应用侧异常处理,并保证不再访问该段内存。尽管CXL 2.0规范的9.9章节对热插拔实施流程给出了简要说明,但实现该功能属于系统级工程,难度极大。
3) DMA设备对CXL协议的支持适配。存储是通过网卡的DMA收发I/O,数据落盘时也要通过磁盘控制器的DMA把内存数据写入到盘中。但是,当前的网卡和磁盘控制器都不支持CXL协议,因此如果将全部缓存数据都放在CXL内存上,则需要在DRAM与CXL内存之间进行1次数据拷贝才能完成存储的I/O流程。若DMA设备能够支持CXL协议,存储系统便可减少1次内存拷贝操作。
CXL是目前性能最好的高速互连技术[47-48],本文提出了一种企业级盘阵存储的CXL全局缓存设计,在浪潮信息存储产品上实现了系统原型。该设计充分利用了CXL内存的分离式和高性能特征,将存储写缓存从控制器内解耦到外部,具有以下技术优势:控制器故障不会导致缓存数据丢失,可以实现真正的控制器横向扩展、n个控制器同时故障n−1、性能大幅提升。4.4节中的分析证明了这是当前可用性最高的全局缓存技术。实验数据表明,CXL全局缓存比RDMA全局缓存的写性能提升了38%。
下一步工作有3个方面:1)把CXL内存和交换机集成到存储内部。2)解决CXL设备的热插拔问题,确保CXL设备突然被拔出时,存储I/O不中断,提升系统可用性。3)把存储的自精简、重删元数据从SSD迁移到大容量的CXL内存上,把集群数据放在CXL内存中,提高存储的性能。
作者贡献声明:李强负责方案整体设计、实验环境搭建和测试工作,并撰写论文;刘涛制定研究方向,指导整体研究路线;张雪庆完成了可用性分析工作,并提出论文修改意见;孟宪伟负责内存管理模块的开发工作;孙明刚负责缓存模块的软件开发工作;李博乐完善论文内容,负责论文结构的讨论与修改;陈曦完成了CXL JBOM的设计和开发工作。

参考文献 引证文献
排序方式:
1
CXL. CXL® specification - compute express[EB/OL]. [2025-05-24]. https://computeexpresslink.org/cxl-specification
2
舒继武, 数据存储架构与技术[M]. 北京: 人民邮电出版社, 2023: 50−58
Shu Jiwu. Data Storage Architecture and Technology[M]. Beijing: Posts & Telecom Press, 2023: 50−58 (in Chinese)
3
吴晨涛, 信息存储与IT管理[M]. 北京: 人民邮电出版社, 2015
Wu Chentao. Information Storage and IT Management[M]. Beijing: Posts & Telecom Press, 2015 (in Chinese)
4
浪潮信息. 浪潮信息高端存储系统 HF18000G6-CGF[EB/OL]. [2025-05-24]. https://www.ieisystem.com/product/storage/15598.html
IEIT. IEIT high-end storage system HF18000G6-CGF[EB/OL]. [2025-05-24]. https://www.ieisystem.com/product/storage/15598.html (in Chinese)
5
华为. OceanStor Dorado V6 高端全闪存存储产品技术白皮书[EB/OL]. [2025-05-24]. https: //e.huawei.com/cn/material/storage/all-flash-storage/42b9aab381c743b1b68c73db12931e3d
Huawei. OceanStor Dorado V6 high-end all-flash storage product technical white paper[EB/OL]. [2025-05-24]. https://e.huawei.com/cn/material/storage/all-flash-storage/42b9aab381c743b1b68c73db12931e3d (in Chinese)
6
EMC. A detailed overview of the next generation PowerMax platform[EB/OL]. [2025-05-24]. https://infohub.delltechnologies.com/static/media/client/7phukh/DAM_3ddff5ea-f554-4502-9304-59a014d9daea.pdf
7
VastData. White paper: The VAST data platform[EB/OL]. [2025-05-24]. https://www.vastdata.com/whitepaper
8
HPE. HPE Alletra storage MP B10000 architecture[EB/OL]. [2025-05-24]. https://www.hpe.com/psnow/doc/a50008302enw?jumpId=in_psnow_ee8f0a3a-72f8-47de-bf38-7b94e0de4c77_gaiw
9
Montage. Montage technology delivers the world’s first CXL™ memory eXpander controller[EB/OL]. [2025-05-24]. https://www.montage-tech.com/Press_Releases/20220506
10
Roland K. Montage technology CXL add in card[EB/OL]. [2025-05-24]. https://drive.google.com/file/d/1bpVLz533F2YTVqN3mIqtr--ngiR35lBe/view?usp=share_link
11
Samsung. Samsung electronics introduces industry’s first 512GB CXL memory module[EB/OL]. (2022-05-10)[2025-05-24]. https://semiconductor.samsung.com/us/news-events/news/samsung-electronics-introduces-industrys-first-512gb-cxl-memory-module/
12
Samsung. CMM-B: Memory pooling appliance for rack computing[EB/OL]. (2022-05-10)[2025-05-24]. https://semiconductor.samsung.com/us/about-us/us-office/us-r-and-d-labs/memory-labs/cxl-memory-module-box/
13
Gao Xianyang, Chen Xi, Wang Yun. Investigating the memory extension and disaggregated memory pooling system[C]//Proc of the 4th Int Conf on Electronic Information Engineering and Computer Science. Piscataway, NJ: IEEE, 2024: 667−671
14
XConn. XConn technologies launches the industry’s first and only hybrid CXL 2.0 and PCIe gen 5 switch[EB/OL]. (2023-08-08)[2025-05-24]. https://www.prnewswire.com/news-releases/xconn-technologies-launches-the-industrys-first-and-only-hybrid-cxl-2-0-and-pcie-gen-5-switch-301895313.html?tc=eml_cleartime
15
Samsung. Samsung memory-semantic CXL SSD[EB/OL]. (2022-08-18)[2025-05-24]. https://www.servethehome.com/samsung-memory-semantic-cxl-ssd-at-fms-2022-powered-by-amd-xilinx/
16
Jung M. Hello bytes, bye blocks: PCIe storage meets compute express link for memory expansion (CXL-SSD)[C]//Proc of the 14th ACM Workshop on Hot Topics in Storage and File Systems. New York: ACM, 2022: 45−51
17
Memverge. Memory pooling and sharing solution[EB/OL]. (2023-08-08)[2025-05-24]. https://memverge.com/samsung-memverge-h3-platform-and-xconn-demonstrate-memory-pooling-and-sharing-for-endless-memory/
18
Li Huaicheng, Berger D S, Hsu L, et al. Pond: CXL-based memory pooling systems for cloud platforms[C]//Proc of the 28th ACM Int Conf on Architectural Support for Programming Languages and Operating Systems. New York: ACM, 2023: 574−587
19
Berger D S, Ernst D, Li H, et al. Design tradeoffs in CXL-based memory pools for public cloud platforms[J]. IEEE Micro, 2023, 43(2): 30−38
20
Gouk D, Lee S, Kwon M, et al. Direct access, high-performance memory disaggregation with directCXL[C]//Proc of USENIX Annual Technical Conf 2022. Berkeley, CA: USENIX Association, 2022: 287−294
21
Sharma D D. Compute express link (CXL): Enabling heterogeneous data-centric computing with heterogeneous memory hierarchy[J]. IEEE Micro, 2022, 43(2): 99−109
22
Gouk D, Kwon M, Bae H, et al. Memory pooling with CXL[J]. IEEE Micro, 2023, 43(2): 48−57
23
Liu Jinshu, Hadian H, Xu Hanchen, et al. Systematic CXL memory characterization and performance analysis at scale[C]// Proc of the 30th ACM Int Conf on Architectural Support for Programming Languages and Operating Systems. New York: ACM, 2025: 1203−1217
24
Wang Xi, Liu Jie, Wu Jianbo, et al. Exploring and evaluating real-world CXL: Use cases and system adoption[J]. arXiv preprint, arXiv: 2405.14209, 2024
25
Sun Yan, Yuan Yifan, Yu Zeduo, et al. Demystifying CXL memory with genuine CXL-ready systems and devices[C]//Proc of the 56th Annual IEEE/ACM Int Symp on Microarchitecture. New York: ACM, 2023: 105−121
26
Kwon M, Lee S, Jung M. Cache in hand: Expander-driven CXL prefetcher for next generation CXL-SSD[C]//Proc of the 15th ACM Workshop on Hot Topics in Storage and File Systems. New York: ACM, 2023: 24−30
27
Yang Shaopeng, Kim M, Nam S, et al. Overcoming the memory wall with CXL-enabled SSDs[C]//Proc of USENIX Annual Technical Conf 2023. Berkeley, CA: USENIX Association, 2023: 601−617
28
Wang Zhonghua, Guo Yixing, Lu Kai, et al. Rcmp: Reconstructing RDMA-based memory disaggregation via CXL[J]. ACM Transactions on Architecture and Code Optimization, 2024, 21(1): 1−26
29
Wang Chenjiu, He Ke, Fan Ruiqi, et al. CXL over ethernet: A novel FPGA-based memory disaggregation design in data centers[C]//Proc of the 31st IEEE Annual Int Symp on Field-Programmable Custom Computing Machines. Piscataway, NJ: IEEE, 2023: 75−82
30
Kim K, Kim H, So J, et al. SMT: Software-defined memory tiering for heterogeneous computing systems with CXL memory expander[J]. IEEE Micro, 2023, 43(2): 20−29
31
Maruf H A, Wang Hao, Dhanotia A, et al. Tpp: Transparent page placement for CXL-enabled tiered-memory[C]//Proc of the 28th ACM Int Conf on Architectural Support for Programming Languages and Operating Systems. New York: ACM, 2023: 742−755
32
Xiang Lingfeng, Lin Zhen, Deng Weishu, et al. Nomad: Non-exclusive memory tiering via transactional page migration[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2024: 19−35
33
Zhong Yuhong, Berger D S, Waldspurger C, et al. Managing memory tiers with CXL in virtualized environments[C]//Proc of the 18th USENIX Symp on Operating Systems Design and Implementation. Berkeley, CA: USENIX Association, 2024: 37−56
34
Wang Hancheng, Dai Haipeng, Chen S, et al. Rethinking Hash tables: Challenges and opportunities with compute express link (CXL)[C]//Proc of the ACM Turing Award Celebration Conf-China 2024. New York: ACM, 2024: 23−27
35
Huang Wentao, Sha Mo, Lu Mian, et al. Bandwidth expansion via CXL: A pathway to accelerating in-memory analytical processing[C]//Proc of the 16th Int Workshop on Accelerating Analytics and Data Management Systems Using Modern Processor and Storage Architectures. New York: ACM, 2024: 1−12
36
Ahn M, Willhalm T, May N, et al. An examination of CXL memory use cases for in-memory database management systems using SAP HANA[J]. Proceedings of the VLDB Endowment, 2024, 17(12): 3827−3840
37
Gouk D, Kang S, Bae H, et al. Breaking barriers: Expanding GPU memory with sub-two digit nanosecond latency CXL controller//Proc of the 16th ACM Workshop on Hot Topics in Storage and File Systems. New York: ACM, 2024: 108−115
38
Park S S, Kim K S, So J, et al. An LPDDR-based CXL-PNM platform for TCO-efficient inference of transformer-based large language models[C]//Proc of the 30th IEEE Int Symp on High-Performance Computer Architecture. Piscataway, NJ: IEEE, 2024: 970−982
39
Liu Haifeng, Zheng Long, Huang Yu, et al. Enabling efficient large recommendation model training with near CXL memory processing[C]//Proc of the 51st ACM/IEEE Annual Int Symp on Computer Architecture. Piscataway, NJ: IEEE, 2024: 382−395
40
Arif M, Assogba K, Rafique M M, et al. Exploiting CXL-based memory for distributed deep learning[C]//Proc of the 51st Int Conf on Parallel Processing. Piscataway, NJ: IEEE, 2022: 1−11
41
Jang J, Choi H, Bae H, et al. CXL-ANNS: Software-hardware collaborative memory disaggregation and computation for billion-scale approximate nearest neighbor search[C]//Proc of USENIX Annual Technical Conf 2023. Berkeley, CA: USENIX Association, 2023: 585−600
42
Ma Teng, Liu Zheng, Wei Chengkun, et al. HydraRPC: RPC in the CXL era[C]//Proc of USENIX Annual Technical Conf 2024. Berkeley, CA: USENIX Association, 2024: 387−395
43
Reichard G, Luis R R. MTRR(memory type range register) control[EB/OL]. [2025-05-24]. https://www.kernel.org/doc/html/next/x86/mtrr.html
44
Memverge. Gismo: World’s first CXL-based multi-server shared memory[EB/OL]. (2023-05-23)[2025-05-24]. https://memverge.com/memverge-unveils-worlds-first-CXL-based-multi-server-shared-memory-at-isc/
45
Storage Performance Council. Storage performance council[EB/OL]. [2025-05-24]. https://www.storageperformance.org
46
舒继武, 新型存算分离架构技术展望[J]. 中国计算机学会通讯, 2022, 18(11): 53−60
Shu Jiwu. Technical prospects of new compute-storage disaggregation architecture[J]. CCF Communications, 2022, 18(11): 53−60 (in Chinese)
47
舒继武, 陈游旻, 汪庆, 等. 分离式数据中心的存储系统研究进展[J]. 中国科学: 信息科学, 2023, 53(8): 1503−1528
Shu Jiwu, Chen Youmin, Wang Qing, et al. Progress on storage systems for disaggregated data centers[J]. SCIENTIA SINICA Informationis, 2023, 53(8): 1503−1528 (in Chinese)
48
Wang Jing, Li Chao, Wang Taolei, et al. Survey of disaggregated memory: Cross-layer technique insights for next-generation datacenters[J]. arXiv preprint, arXiv: 2503.20275, 2025
49
Yang Xinjun, Zhang Yingqiang, Chen Hao, et al. Unlocking the potential of CXL for disaggregated memory in cloud-native databases[C]//Proc of the 2025 Companion of Int Conf on Management of Data. New York: ACM, 2025: 689−702
50
Intel. Intel memory latency checker (MLC)[EB/OL]. [2025-05-24]. https://www.intel.cn/content/www/cn/zh/download/736633/736634/intel-memory-latency-checker-intel-mlc.html
2026年第63卷第3期
192
79
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202550376
  • 接收时间:2025-06-04
  • 首发时间:2026-11-06
  • 出版时间:2026-03-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2025-06-04
  • 修回日期:2025-11-21
  • 录用日期:2025-11-28
基金
作者信息
    1浪潮电子信息产业股份有限公司 济南 250101
    2华中科技大学计算机科学与技术学院 武汉 430074
    3浪潮(北京)电子信息产业有限公司 北京 100095

通讯作者:

张雪庆()
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202550376
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏