Reliable Main-Memory Storage Systems

可靠的主内存存储系统

• 批准号: 9521386
• 负责人: Willy Zwaenepoel
• 金额: $ 143.99万
• 依托单位: William Marsh Rice University
• 依托单位国家: 美国
• 项目类别: Continuing grant
• 财政年份: 1995
• 资助国家: 美国
• 起止时间: 1995-12-01 至 2001-11-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9521386&HistoricalAwards=false
• 关键词: Reliable; Main; Memory; Storage; Systems

This project investigates the use of non-volatile semiconductor memory (such as flash memory) as a storage system to replace some of the functions currently assigned to disks. A board (eNVy) incorporating a large (128MB) flash memory is being designed at Rice University; concurrently, a log-structured file system (Rio) that can use the flash memory board as a cache between main memory and disk is being developed at the University of Michigan. In later stages of the project these research prototypes will be integrated. Research issues in hardware development include space reclamation (garbage collection), and tradeoffs between memory size, performance, and memory lifetimes. Research issues in file system development include optimization of the sizes of cache and main memory, algorithms for compression of data in the cache, protection of memory from software faults, and exploration of file system index structures. Integration of the prototypes will provide insight into the protability of these techniques to new systems.

该项目研究使用非易失性半导体存储器（例如闪存）作为存储系统来取代目前分配给磁盘的一些功能。 莱斯大学正在设计一款包含大型 (128MB) 闪存的板 (eNVy)；与此同时，密歇根大学正在开发一种日志结构文件系统（Rio），它可以使用闪存板作为主存和磁盘之间的缓存。 在项目的后期阶段，这些研究原型将被整合。 硬件开发中的研究问题包括空间回收（垃圾收集）以及内存大小、性能和内存寿命之间的权衡。 文件系统开发的研究问题包括缓存和主存大小的优化、缓存中数据的压缩算法、保护内存免受软件故障的影响以及文件系统索引结构的探索。 原型的集成将提供对这些技术在新系统中的可移植性的深入了解。