ACM Transactions on Computer Systems

ACM Transactions on Computer Systems is a quarterly journal of the Association for Computing Machinery, devoted to the design, implementation and measured behaviour of computer systems and systems software. Volume 1 appeared in February 1983, and the 2025 Journal Impact Factor stands at 1.9.《ACM 计算机系统汇刊》（ACM Transactions on Computer Systems）是美国计算机协会出版的系统类季刊，关注计算机系统与系统软件的设计、实现与实测表现。第 1 卷于 1983 年 2 月问世，2025 年影响因子为 1.9。

Four decades of volumes hold a small archive: about a dozen citable papers a year, 1,253 citations in the 2025 record and a cited half-life of 17.7 years, numbers that describe a slowly consumed backfile rather than a high-volume outlet.四十余卷积累下来的是一份不大的存档，每年可引用论文十余篇，2025 年累计被引 1253 次，引用半衰期 17.7 年，这些数字对应一份慢热而过刊耐读的刊物，产量并不高。

Scope

• The stated remit covers research and development on the design, implementation and use of computer systems and systems software, and the word system is read broadly rather than tied to one layer of the stack.刊物明列的职责，是计算机系统与系统软件在设计、实现与使用上的研发成果，其中「系统」一词取宽泛含义，不限于软件栈的某一层。
• Operating systems anchor the file, taking in kernel structure, scheduling and memory management, the file systems layered on block devices, and the interfaces through which applications meet hardware.操作系统是这份刊物的根基，涉及内核结构、调度与内存管理，也涉及建立在块设备之上的文件系统，以及应用程序与硬件相接的接口。
• Distributed systems carry equal weight, from remote procedure call and group communication to consensus, replication and the failure detectors on which such protocols rest.分布式系统的分量相当，从远程过程调用与组通信，到共识、复制，以及这类协议所依赖的失效检测。
• Storage work runs from flash translation layers and log-structured designs to persistent memory, deduplication and the durability a file system promises an application.存储方向从闪存转换层与日志结构设计，一直延伸到持久内存与去重，也延伸到文件系统向应用许下的持久性承诺。
• Virtualization and isolation are treated as problems in their own right, covering hypervisors, containers, trusted execution environments and the accounting that keeps tenants apart.虚拟化与隔离被当作独立的系统问题处理，涵盖管理程序、容器与可信执行环境，也涵盖把租户彼此隔开的计量。
• Systems and computer architecture meet in these pages, through cache behaviour, memory hierarchies, accelerators and the operating system support that new hardware demands.系统与体系结构在版面上交汇，涉及缓存行为、存储层次与加速器，以及新硬件所要求的操作系统支持。
• Concurrency and runtime support appear as lock-free data structures, memory models and garbage collection, together with the language mechanisms that make such guarantees usable by ordinary programmers.并发与运行时支持以无锁数据结构、内存模型与垃圾回收的形式出现，也以让这些保证可为普通程序员所用的语言机制出现。
• Optimizing compilers are named in the scope, with papers on code generation, profile-guided transformation and the way compiler choices interact with hardware behaviour.优化编译器写进了范围，刊载代码生成、基于剖面的变换，以及编译器选择与硬件行为相互作用的研究。
• The boundary with networking stays open, so transport design, congestion control and the placement of protocol processing between kernel and user space all find a home.与网络的边界保持开放，传输层设计、拥塞控制，以及协议处理究竟放在内核还是用户态，都能在此落脚。
• Secure systems belong to the remit as well, including isolation primitives, attestation, side-channel measurement and the threat models against which such defences are judged.安全系统同样在范围内，包括隔离原语、远程证明与侧信道测量，也包括用以检验这些防御的威胁模型。
• Performance and power optimization are listed together, a reflection of a field where throughput per watt has become as reportable as latency or instruction count.性能与功耗优化被列在一处，反映出这一领域里每瓦吞吐已经和延迟、指令数一样值得报告。
• Machine learning systems entered the scope through training pipelines and serving stacks, along with the memory and interconnect pressure that large models place on a single machine.机器学习系统经由训练流水线与服务栈进入范围，也经由大模型对单机内存与互连造成的压力进入范围。
• Cloud and datacenter systems, embedded and mobile platforms, dependable computing and system management fill out the list of topics printed on the submission page.云与数据中心系统、嵌入式与移动平台，加上可信计算与系统管理，共同补足了投稿页上明列的主题。
• Articles divide into two kinds by design: those that present new techniques and concepts, and those that report experience with and measurement of systems already running in the field.文章按设计分为两类，一类提出新的技术与概念，一类报告已在现场运行的系统的经验与实测。
• Research papers appear in long and short form, and a technical communications track allows comment on a technical topic or on a paper the journal has already published.研究论文分长篇与短篇两种，另设技术通讯一栏，可就某一技术专题或本刊已发表的论文提出评论。
• Emphasis rests on insight useful to the people who design, build and operate systems, which leaves room for a clean experiment that ends in a negative result.重心落在对系统设计者、实现者与运行者有用的洞见上，实验做得干净而结果为否的工作，同样有被接受的空间。

History

• Volume 1, number 1 is dated February 1983 and came out under the ACM imprint, with the society acting as publisher in its own right, an arrangement that has never changed since.第 1 卷第 1 期标注 1983 年 2 月，由 ACM 自行印行，协会既是主办方也是出版方，这一格局此后没有变过。
• That issue opens with a two-page editor's introduction signed by Anita Jones, who set the pattern of long technical papers with measured evaluation that the title still follows.那一期以两页的编者导言开卷，署名 Anita Jones，她定下的长篇技术论文加实测评估的路子，刊物一直沿用。^[1]
• Systems research in 1983 published mostly through conference proceedings, and a quarterly transaction gave a paper the room to describe an implementation at a depth a referee could check.1983 年的系统研究大多经会议录发表，一份季刊汇刊让论文有篇幅把实现讲到评审能够核对的细度。
• Volume numbers have advanced one per calendar year without a gap, so volume 44 belongs to 2026 and nobody needs a conversion table to date a paper.卷号每年递增一卷，中间没有断档，2026 年对应第 44 卷，给论文断代不必查换算表。
• From 1983 through volume 36 four issues made a volume, and the papers of that run are still the ones reference lists cite most, Sprite's network file system caching among them.从 1983 年到第 36 卷，每卷四期，那一批论文至今仍最常出现在参考文献里，Sprite 的网络文件系统缓存研究就在其中。^[2]
• Crossref holds 623 records for the title, and the earliest of them are deposited with one continuous pagination per issue, the layout of a bound quarterly.Crossref 为这份刊物收录 623 条记录，最早的一批按每期一套连续页码登记，是合订季刊的排法。
• The quarterly rhythm loosened after that. Volume 37 in 2019 is deposited as a single issue numbered 1-4, and later volumes follow with two combined issues or one for the whole year.季刊的节奏此后放宽。2019 年的第 37 卷整卷登记为编号 1-4 的一期，随后几卷或并为两期，或整年只留一期。
• Numbering returned in 2025, when volume 43 splits into 1-2, 3 and 4, and volume 44 in 2026 carries all four issues with some thirty papers among them.2025 年编号恢复，第 43 卷拆成 1-2、3 与 4 三期，2026 年的第 44 卷四期齐备，合计约三十篇论文。
• Accepted work appears as soon as it is typeset, each item with a DOI and an article number, and the issue is assembled afterwards. Eleven records dated 2026 still wait for a volume.录用的论文排版完成即上线，各带 DOI 与文章号，期次随后汇编。2026 年仍有十一条记录尚未归卷。
• Every DOI begins with 10.1145, the prefix ACM uses across its entire list, so a paper here sits beside SIGCOMM and ASPLOS work under a single registrant.所有 DOI 都以 10.1145 开头，这是 ACM 全社共用的前缀，本刊论文因此与 SIGCOMM 及 ASPLOS 的工作同属一个注册方。
• Ownership never moved. Where several systems journals passed to commercial publishers, this one stayed with the society, and distribution shifted to the ACM Digital Library in the late 1990s.所有权始终没有转移。好几种系统刊物后来转到了商业出版集团手里，这一份留在学会内部，1990 年代后期发行改由 ACM Digital Library 承担。
• OpenTOC opens each issue's table of contents and the links behind it at no cost, which is how many readers first meet a paper before their library pays for the PDF.OpenTOC 免费开放每期目录与目录后的链接，不少读者先在这里遇到论文，再由所在图书馆付费取全文。
• The journal is hybrid rather than fully open: subscriptions carry the record, and authors under a mandate may pay for gold open access, which the 2025 data put at 32 percent of output.刊物走混合模式，订阅支撑正刊，有开放要求的作者可付费选择金色开放，2025 年的统计把这一比例记作 32%。
• Impact factors over five years read 1.75 for 2021, 1.692 for 2022, 2.0 for 2023, 1.8 for 2024 and 1.9 for 2025, against a five-year figure of 2.5.近五年影响因子依次为 2021 年 1.75，2022 年 1.692，2023 年 2.0，2024 年 1.8，2025 年 1.9，五年影响因子为 2.5。
• The 2025 placement is 73rd of 146 in computer science, theory and methods, a Q2 slot, with a Journal Citation Indicator of 0.49 and a percentile of 50.3.2025 年的位置是计算机科学理论与方法类目 146 种中的第 73 位，属 Q2，期刊引文指标 0.49，百分位 50.3。
• Twelve citable papers a year and 1,253 citations in total describe the size of the operation, while a 17.7-year cited half-life says the older volumes still earn their keep.每年 12 篇可引用论文与 1253 次累计被引，说明了这份刊物的规模，17.7 年的引用半衰期则说明旧卷仍在挣得引用。
• Subject matter shifted with the field. Where early volumes argued about kernels and remote procedure call, recent ones carry enclaves, disaggregated memory and the systems that train and serve large models.题材随学科移动。早期各卷争论内核与远程过程调用，近年的版面上则是安全飞地、分离式内存，以及训练和部署大模型的系统。

Editors

• Sam H. Noh of Virginia Tech and Robert van Renesse of Cornell University serve as co-editors-in-chief, appointed for a term that runs from May 2024 to April 2027.主编由 Virginia Tech 的 Sam H. Noh 与 Cornell University 的 Robert van Renesse 共同担任，任期自 2024 年 5 月至 2027 年 4 月。
• The pairing mirrors the two halves of the field. Noh works on storage and file systems, van Renesse on distributed computing and fault tolerance, and either of them may take charge of a submission.两人的搭配映出领域的两半，Noh 研究存储与文件系统，van Renesse 研究分布式计算与容错，来稿可由其中任何一位接手。
• Below them sits a group of associate editors drawn from universities and industrial laboratories across North America, Europe and Asia, each taking the submissions that fall inside one area of systems work.其下是一组副编辑，来自北美、欧洲与亚洲的高校与工业实验室，各自接手落在自己那块系统领域之内的来稿。
• Manuscripts go through ScholarOne at mc.manuscriptcentral.com/acm/tocs, and the journal asks for a complete paper rather than a pre-submission enquiry.稿件经 ScholarOne 递交，入口为 mc.manuscriptcentral.com/acm/tocs，刊物要的是完整论文，不受理投稿前的问询。
• Referees are chosen by the handling editor from outside the editorial office, and a systems paper is weighed on its implementation, its experimental setup and the baselines it compares against.评审由经手编辑在编辑部之外约请，系统类论文按实现、实验设置与对照基线来衡量。
• Review runs slowly by the standards of this corner of computing. Long papers commonly see two rounds, and the dozen articles a year follow from that filter rather than from a shortage of submissions.按这一领域的标准衡量，评审偏慢。长文常经两轮修改，全年十余篇的产量来自这道筛子，与来稿不足无关。
• The editorial preference is stated in the scope and visible in the pages: a system that ran under a real workload counts for more than a simulation of one that might.偏好写在范围里，也显在版面上，在真实负载下跑过的系统，比一份假想系统的模拟更受看重。
• Where a claim rests on measurement, editors tend to ask for the raw numbers, the machine configuration and the version of the code that produced them.结论建立在测量之上的稿件，编辑往往要求交出原始数据、机器配置与产生这些数字的代码版本。
• ACM's artifact review and badging programme gives authors a route to have code and data examined, and systems submissions increasingly arrive with a repository attached.ACM 的制品评审与徽章制度为作者提供了送检代码与数据的通道，系统类稿件附带仓库的做法渐多。
• Comment on published work has a channel of its own, since a technical communication may dispute a result and the original authors are offered space to answer.对已发表工作的意见另有通道，技术通讯可以就某一结果提出异议，原作者获得答复的版面。
• Surveys are rare. The space is kept for papers that build or measure something, so anyone proposing a review has to argue why the field needs it now.综述少见。版面留给做出东西或测出数据的论文，想写综述的人须说明这一领域此刻为何需要它。
• Corrections and retractions follow ACM policy, appear as items of their own and stay linked to the version of record they amend.更正与撤稿依 ACM 的政策处理，作为独立条目刊出，并与它们所修正的正式版本保持链接。
• Authors choose between ACM copyright and a CC BY licence at acceptance, and that choice decides whether the paper is counted as gold open access.录用时作者在 ACM 版权与 CC BY 许可之间择一，这一选择决定论文是否计入金色开放。