Citation Context

...ies on previous work in areas such as scalable cache consistency (DASH [Leno90] and Alewife [Chai91]), cooperative caching, disk striping (RAID and Zebra), and log structured file systems (Sprite LFS =-=[Rose92]-=- and BSD LFS [Selt93]). Second, in addition to borrowing techniques developed in other projects, we have refined them to work well in our serverless system. For instance, we have transformed DASH’s sc...

Citation Context

... continue to improve, then writes will eventually dominate file system performance and new approaches, such as longer writeback intervals, non-volatile cache memories, and log-structured file systems =-=[15]-=-, will become attractive. We found that many users access file data in a way that assumes cache consistency among workstations, and that they will be inconvenienced on a daily basis if full consistenc...

Citation Context

...e storage capacity of the disk array. Second, when written data is replaced in the cache, it is not written back to the same place on disk. In a manner much like Berkeley’s Log Structured File System =-=[Rosenblum91]-=-, data is written opportunistically to disk in large track-sized transfer units, reducing random access latencies and performing adaptive load balancing. And third, the cache makes it possible to tran...

Citation Context

...pani’s logging is an application of techniques first developed for databases [2] and later used in several other log-based file systems [9, 11, 16, 18]. Frangipani is not a log-structured file system =-=[32]-=-; it does not keep all its data in the log, instead maintaining conventional on-disk data structures, with a small log as an adjunct to provide improved performance and failure atomicity. Unlike the o...

Citation Context

...is a network file system architecture designed to provide both high performance and high availability. This is accomplished by incorporating ideas from log-structured file systems, such as Sprite LFS =-=[Rosenblum91]-=-, and redundant arrays of inexpensive disks (RAID) [Patterson88] into a network file system. From log-structured file systems Zebra borrows the idea that small, independent writes to the storage subsy...

Citation Context

...ly. Ivy's primary contribution relative to these systems is that it uses peer-topeer storage to build a read/write file system that multiple users can share. 7.1 Log-structured File System Sprite LFS =-=[31]-=- represents a file system as a log of operations, along with a snapshot of i-number to i-node location mappings. LFS uses a single log managed by a single server in order to to speed up small write pe...

Citation Context

...y the cache absorbs read requests. In particular, we are interested in how effective caches are at reducing disk seeks and how caching affects the balance between disk reads and writes. In this section, we examine the effect of cache size on read misses. We find that even relatively small caches absorb most read traffic, but there are diminishing returns to using larger caches. We also examine how caching affects the ratio of disk reads to disk writes. In 1992, Rosenblum and Ousterhout claimed that large caches would avert most disk reads, so file system layout should optimize for disk writes [Rose92]. We FIGURE 4. Read Bandwidth versus Cache Size. This graph shows the percentage of all block read requests that miss the cache versus cache size. The block size used by the cache simulator is 4KB. The cache was warmed with a day of traces before generating results. 0 20 40 60 80 100 1 2 4 8 16 32 64 128 256 M is s R at e Cache Size (MB) ins res web nt find that the read to write ratio depends not only on the cache size, but also on the write delay and workload. Finally, we examine how well caching works for memory-mapped files. We find that because a small number of files tend to be memory-ma...