A case for redundant arrays of inexpensive disks (RAID)
SIGMOD '88 Proceedings of the 1988 ACM SIGMOD international conference on Management of data
Inside the Windows NT File System
Inside the Windows NT File System
Forecasting Disk Resource Requirements for a Usenet Server
LISA '93 Proceedings of the 7th USENIX conference on System administration
File system logging versus clustering: a performance comparison
TCON'95 Proceedings of the USENIX 1995 Technical Conference Proceedings
The Cyclic News Filesystem: Getting INN To Do More With Less
LISA '97 Proceedings of the 11th USENIX conference on System administration
Semantically-Smart Disk Systems
FAST '03 Proceedings of the 2nd USENIX Conference on File and Storage Technologies
Journaling versus soft updates: asynchronous meta-data protection in file systems
ATEC '00 Proceedings of the annual conference on USENIX Annual Technical Conference
A nine year study of file system and storage benchmarking
ACM Transactions on Storage (TOS)
Semantically-smart disk systems
FAST'03 Proceedings of the 2nd USENIX conference on File and storage technologies
Hi-index | 0.00 |
Usenet volume has been growing exponentially for many years; this growth places ever-increasing demands on the resources of a netnews server, particularly disk space - and file system performance - for the article spool area. Keeping up with this demand became substantially more difficult when it could no longer be satisfied by a single disk, and since netnews is incidental to SLAC's research mission, we wanted to find a solution that could easily scale to meet future growth while requiring minimal system administration effort. In the process of evaluating the various solutions that were proposed, we developed benchmarks to measure performance for this specialized application, and were surprised to find that some of our beliefs and intuition were not supported by the facts. The alternatives considered by SLAC are described, as are the benchmarks developed to evaluate the alternatives, and the results of those benchmarks. While netnews is the application we examined, our experience will hopefully provide inspiration for others to more carefully evaluate their applications instead of using stock benchmarks that may not correlate well with the intended use. Our results may also break down some biases and encourage the reader to consider alternatives which might otherwise have been ignored.