{"title":"LLLFS: A Low-latency Library File System for Persistent Memory","authors":"K. Matsuzawa, Takahiro Shinagawa","doi":"10.1109/UCC56403.2022.00062","DOIUrl":null,"url":null,"abstract":"Persistent memory (PM) is a promising storage medium for reducing file access latency, which is critical for modern applications. Unfortunately, traditional storage stacks designed for slow media significantly impair the low-latency access capability of PM. We propose LLLFS, a low-latency library file system for PM. Based on the observation that many modern applications access files under their directories exclusively, LLLFS stores both file data and metadata in a process-local file system on PM so that files are fully accessible in user space without system calls and synchronization. LLLFS keeps POSIX compatibility while allowing application-specific parameter tuning. Experimental results demonstrated that LLLFS outperformed conventional and PM-optimized file systems by 56.0% and 37.7%, respectively, in database update operations.","PeriodicalId":203244,"journal":{"name":"2022 IEEE/ACM 15th International Conference on Utility and Cloud Computing (UCC)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE/ACM 15th International Conference on Utility and Cloud Computing (UCC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/UCC56403.2022.00062","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Persistent memory (PM) is a promising storage medium for reducing file access latency, which is critical for modern applications. Unfortunately, traditional storage stacks designed for slow media significantly impair the low-latency access capability of PM. We propose LLLFS, a low-latency library file system for PM. Based on the observation that many modern applications access files under their directories exclusively, LLLFS stores both file data and metadata in a process-local file system on PM so that files are fully accessible in user space without system calls and synchronization. LLLFS keeps POSIX compatibility while allowing application-specific parameter tuning. Experimental results demonstrated that LLLFS outperformed conventional and PM-optimized file systems by 56.0% and 37.7%, respectively, in database update operations.