Scale and performance in a distributed file system
ACM Transactions on Computer Systems (TOCS)
Caching in the Sprite network file system
ACM Transactions on Computer Systems (TOCS)
Coda: A Highly Available File System for a Distributed Workstation Environment
IEEE Transactions on Computers
Measurements of a distributed file system
SOSP '91 Proceedings of the thirteenth ACM symposium on Operating systems principles
The LRU-K page replacement algorithm for database disk buffering
SIGMOD '93 Proceedings of the 1993 ACM SIGMOD international conference on Management of data
Prefetching over a network: early experience with CTIP
ACM SIGMETRICS Performance Evaluation Review
Daisy: virtual-disk hiearchical storage manager
ACM SIGMETRICS Performance Evaluation Review
A trace-driven analysis of the UNIX 4.2 BSD file system
Proceedings of the tenth ACM symposium on Operating systems principles
Long term file migration: development and evaluation of algorithms
Communications of the ACM
2Q: A Low Overhead High Performance Buffer Management Replacement Algorithm
VLDB '94 Proceedings of the 20th International Conference on Very Large Data Bases
Analysis of Long-Term UNIX File Access Patterns for Application
Analysis of Long-Term UNIX File Access Patterns for Application
Long-term unix file system activity and the efficacy of automatic file migration
Long-term unix file system activity and the efficacy of automatic file migration
Analysis of Long Term File Reference Patterns for Application to File Migration Algorithms
IEEE Transactions on Software Engineering
Hi-index | 0.01 |
Caching jukeboxes combine the low media cost of CD and DVD with the higher performance of magnetic disk drives. We propose and evaluate the combination of aggressive file prefetching based on the a priori affinity among a file's blocks, with block-level usage tracking and removal. This apparently-inconsistent approach takes into account the usage time constants rather than merely the relative performance of two storage levels: the expected lifetime of a block in the large disk cache is long, rendering the fine-grain usage information much more meaningful than in higher levels of the memory hierarchy. Preliminary measurements carried out on our IntelliJuke prototype, along with comparative simulations, confirm the direct cost-performance benefits of our approach. One side benefit is increased effective mechanical reliability of the jukebox. Finally, this work motivates the support of sparse files on disk and illustrates the benefits of making additional information available to storage systems.