Non-volatile memory for fast, reliable file systems
ASPLOS V Proceedings of the fifth international conference on Architectural support for programming languages and operating systems
Management of Partially Safe Buffers
IEEE Transactions on Computers
Conquest: Better Performance Through a Disk/Persistent-RAM Hybrid File System
ATEC '02 Proceedings of the General Track of the annual conference on USENIX Annual Technical Conference
HeRMES: High-Performance Reliable MRAM-Enabled Storage
HOTOS '01 Proceedings of the Eighth Workshop on Hot Topics in Operating Systems
Second-Level Buffer Cache Management
IEEE Transactions on Parallel and Distributed Systems
MRAMFS: A Compressing File System for Non-Volatile RAM
MASCOTS '04 Proceedings of the The IEEE Computer Society's 12th Annual International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunications Systems
NBM: an efficient cache replacement algorithm for nonvolatile buffer caches
ACS'08 Proceedings of the 8th conference on Applied computer scince
Impact of NVRAM write cache for file system metadata on I/O performance in embedded systems
Proceedings of the 2009 ACM symposium on Applied Computing
Hi-index | 0.00 |
Nonvolatile RAM (NVRAM) technology is advancing rapidly with 1-2Mb capacity single-chip prototypes becoming available from major semiconductor companies. We will soon see NVRAM become an everyday component of our commodity computers. This paper explores the use of NVRAM as part of the buffer cache. A nonvolatile buffer cache provides a computer system with a means to maintain complete consistency as well as improved performance. The results of this paper can be summarized as follows. First, we show that the hit ratio that has been a commonly used metric to measure buffer cache performance is no longer adequate for caches with NVRAM. Instead of the hit ratio, we need to count the number of disk accesses to assess user perceived cache performance. Second, we show that because of this change in performance metric, when managing a buffer cache with NVRAM, one can do better than when using the MIN replacement algorithm mainly by distinguishing read and write operations. With this, we show that there is room for improvement in efficiently handling caches with NVRAM. Finally, based on these findings, we propose a simple and practical buffer management technique that improves on using the LRU algorithm.